]>
Commit | Line | Data |
---|---|---|
11fdf7f2 | 1 | #!/usr/bin/env bash |
7c673cae FG |
2 | # |
3 | # Copyright (C) 2013,2014 Cloudwatt <libre.licensing@cloudwatt.com> | |
4 | # Copyright (C) 2014,2015 Red Hat <contact@redhat.com> | |
5 | # Copyright (C) 2014 Federico Gimenez <fgimenez@coit.es> | |
6 | # | |
7 | # Author: Loic Dachary <loic@dachary.org> | |
8 | # Author: Federico Gimenez <fgimenez@coit.es> | |
9 | # | |
10 | # This program is free software; you can redistribute it and/or modify | |
11 | # it under the terms of the GNU Library Public License as published by | |
12 | # the Free Software Foundation; either version 2, or (at your option) | |
13 | # any later version. | |
14 | # | |
15 | # This program is distributed in the hope that it will be useful, | |
16 | # but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
18 | # GNU Library Public License for more details. | |
19 | # | |
20 | TIMEOUT=300 | |
11fdf7f2 TL |
21 | WAIT_FOR_CLEAN_TIMEOUT=90 |
22 | MAX_TIMEOUT=15 | |
7c673cae | 23 | PG_NUM=4 |
11fdf7f2 | 24 | TMPDIR=${TMPDIR:-/tmp} |
f64942e4 AA |
25 | CEPH_BUILD_VIRTUALENV=${TMPDIR} |
26 | TESTDIR=${TESTDIR:-${TMPDIR}} | |
7c673cae FG |
27 | |
28 | if type xmlstarlet > /dev/null 2>&1; then | |
29 | XMLSTARLET=xmlstarlet | |
30 | elif type xml > /dev/null 2>&1; then | |
31 | XMLSTARLET=xml | |
32 | else | |
33 | echo "Missing xmlstarlet binary!" | |
34 | exit 1 | |
35 | fi | |
31f18b77 | 36 | |
7c673cae FG |
37 | if [ `uname` = FreeBSD ]; then |
38 | SED=gsed | |
f64942e4 | 39 | AWK=gawk |
31f18b77 | 40 | DIFFCOLOPTS="" |
b5b8bbf5 | 41 | KERNCORE="kern.corefile" |
7c673cae FG |
42 | else |
43 | SED=sed | |
f64942e4 | 44 | AWK=awk |
31f18b77 | 45 | termwidth=$(stty -a | head -1 | sed -e 's/.*columns \([0-9]*\).*/\1/') |
c07f9fc5 FG |
46 | if [ -n "$termwidth" -a "$termwidth" != "0" ]; then |
47 | termwidth="-W ${termwidth}" | |
31f18b77 FG |
48 | fi |
49 | DIFFCOLOPTS="-y $termwidth" | |
b5b8bbf5 | 50 | KERNCORE="kernel.core_pattern" |
31f18b77 | 51 | fi |
7c673cae | 52 | |
c07f9fc5 | 53 | EXTRA_OPTS="" |
c07f9fc5 | 54 | |
7c673cae FG |
55 | #! @file ceph-helpers.sh |
56 | # @brief Toolbox to manage Ceph cluster dedicated to testing | |
57 | # | |
58 | # Example use case: | |
59 | # | |
60 | # ~~~~~~~~~~~~~~~~{.sh} | |
61 | # source ceph-helpers.sh | |
62 | # | |
63 | # function mytest() { | |
64 | # # cleanup leftovers and reset mydir | |
65 | # setup mydir | |
66 | # # create a cluster with one monitor and three osds | |
67 | # run_mon mydir a | |
68 | # run_osd mydir 0 | |
69 | # run_osd mydir 2 | |
70 | # run_osd mydir 3 | |
71 | # # put and get an object | |
72 | # rados --pool rbd put GROUP /etc/group | |
73 | # rados --pool rbd get GROUP /tmp/GROUP | |
74 | # # stop the cluster and cleanup the directory | |
75 | # teardown mydir | |
76 | # } | |
77 | # ~~~~~~~~~~~~~~~~ | |
78 | # | |
79 | # The focus is on simplicity and efficiency, in the context of | |
80 | # functional tests. The output is intentionally very verbose | |
81 | # and functions return as soon as an error is found. The caller | |
82 | # is also expected to abort on the first error so that debugging | |
83 | # can be done by looking at the end of the output. | |
84 | # | |
85 | # Each function is documented, implemented and tested independently. | |
86 | # When modifying a helper, the test and the documentation are | |
87 | # expected to be updated and it is easier of they are collocated. A | |
88 | # test for a given function can be run with | |
89 | # | |
90 | # ~~~~~~~~~~~~~~~~{.sh} | |
91 | # ceph-helpers.sh TESTS test_get_osds | |
92 | # ~~~~~~~~~~~~~~~~ | |
93 | # | |
94 | # and all the tests (i.e. all functions matching test_*) are run | |
95 | # with: | |
96 | # | |
97 | # ~~~~~~~~~~~~~~~~{.sh} | |
98 | # ceph-helpers.sh TESTS | |
99 | # ~~~~~~~~~~~~~~~~ | |
100 | # | |
101 | # A test function takes a single argument : the directory dedicated | |
102 | # to the tests. It is expected to not create any file outside of this | |
103 | # directory and remove it entirely when it completes successfully. | |
104 | # | |
105 | ||
106 | ||
c07f9fc5 FG |
107 | function get_asok_dir() { |
108 | if [ -n "$CEPH_ASOK_DIR" ]; then | |
109 | echo "$CEPH_ASOK_DIR" | |
110 | else | |
111 | echo ${TMPDIR:-/tmp}/ceph-asok.$$ | |
112 | fi | |
113 | } | |
114 | ||
115 | function get_asok_path() { | |
116 | local name=$1 | |
117 | if [ -n "$name" ]; then | |
118 | echo $(get_asok_dir)/ceph-$name.asok | |
119 | else | |
120 | echo $(get_asok_dir)/\$cluster-\$name.asok | |
121 | fi | |
122 | } | |
7c673cae FG |
123 | ## |
124 | # Cleanup any leftovers found in **dir** via **teardown** | |
125 | # and reset **dir** as an empty environment. | |
126 | # | |
127 | # @param dir path name of the environment | |
128 | # @return 0 on success, 1 on error | |
129 | # | |
130 | function setup() { | |
131 | local dir=$1 | |
132 | teardown $dir || return 1 | |
133 | mkdir -p $dir | |
c07f9fc5 | 134 | mkdir -p $(get_asok_dir) |
11fdf7f2 TL |
135 | if [ $(ulimit -n) -le 1024 ]; then |
136 | ulimit -n 4096 || return 1 | |
137 | fi | |
138 | if [ -z "$LOCALRUN" ]; then | |
139 | trap "teardown $dir 1" TERM HUP INT | |
140 | fi | |
7c673cae FG |
141 | } |
142 | ||
143 | function test_setup() { | |
144 | local dir=$dir | |
145 | setup $dir || return 1 | |
146 | test -d $dir || return 1 | |
147 | setup $dir || return 1 | |
148 | test -d $dir || return 1 | |
149 | teardown $dir | |
150 | } | |
151 | ||
152 | ####################################################################### | |
153 | ||
154 | ## | |
155 | # Kill all daemons for which a .pid file exists in **dir** and remove | |
156 | # **dir**. If the file system in which **dir** is btrfs, delete all | |
157 | # subvolumes that relate to it. | |
158 | # | |
159 | # @param dir path name of the environment | |
1adf2230 | 160 | # @param dumplogs pass "1" to dump logs otherwise it will only if cores found |
7c673cae FG |
161 | # @return 0 on success, 1 on error |
162 | # | |
163 | function teardown() { | |
164 | local dir=$1 | |
b5b8bbf5 | 165 | local dumplogs=$2 |
7c673cae FG |
166 | kill_daemons $dir KILL |
167 | if [ `uname` != FreeBSD ] \ | |
168 | && [ $(stat -f -c '%T' .) == "btrfs" ]; then | |
169 | __teardown_btrfs $dir | |
170 | fi | |
b5b8bbf5 FG |
171 | local cores="no" |
172 | local pattern="$(sysctl -n $KERNCORE)" | |
173 | # See if we have apport core handling | |
174 | if [ "${pattern:0:1}" = "|" ]; then | |
175 | # TODO: Where can we get the dumps? | |
176 | # Not sure where the dumps really are so this will look in the CWD | |
177 | pattern="" | |
178 | fi | |
179 | # Local we start with core and teuthology ends with core | |
1adf2230 | 180 | if ls $(dirname "$pattern") | grep -q '^core\|core$' ; then |
b5b8bbf5 FG |
181 | cores="yes" |
182 | if [ -n "$LOCALRUN" ]; then | |
183 | mkdir /tmp/cores.$$ 2> /dev/null || true | |
184 | for i in $(ls $(dirname $(sysctl -n $KERNCORE)) | grep '^core\|core$'); do | |
185 | mv $i /tmp/cores.$$ | |
186 | done | |
187 | fi | |
188 | fi | |
189 | if [ "$cores" = "yes" -o "$dumplogs" = "1" ]; then | |
1adf2230 AA |
190 | if [ -n "$LOCALRUN" ]; then |
191 | display_logs $dir | |
192 | else | |
193 | # Move logs to where Teuthology will archive it | |
194 | mkdir -p $TESTDIR/archive/log | |
195 | mv $dir/*.log $TESTDIR/archive/log | |
196 | fi | |
b5b8bbf5 | 197 | fi |
7c673cae | 198 | rm -fr $dir |
c07f9fc5 | 199 | rm -rf $(get_asok_dir) |
b5b8bbf5 FG |
200 | if [ "$cores" = "yes" ]; then |
201 | echo "ERROR: Failure due to cores found" | |
202 | if [ -n "$LOCALRUN" ]; then | |
203 | echo "Find saved core files in /tmp/cores.$$" | |
204 | fi | |
205 | return 1 | |
206 | fi | |
207 | return 0 | |
7c673cae FG |
208 | } |
209 | ||
210 | function __teardown_btrfs() { | |
211 | local btrfs_base_dir=$1 | |
f64942e4 AA |
212 | local btrfs_root=$(df -P . | tail -1 | $AWK '{print $NF}') |
213 | local btrfs_dirs=$(cd $btrfs_base_dir; sudo btrfs subvolume list -t . | $AWK '/^[0-9]/ {print $4}' | grep "$btrfs_base_dir/$btrfs_dir") | |
7c673cae FG |
214 | for subvolume in $btrfs_dirs; do |
215 | sudo btrfs subvolume delete $btrfs_root/$subvolume | |
216 | done | |
217 | } | |
218 | ||
219 | function test_teardown() { | |
220 | local dir=$dir | |
221 | setup $dir || return 1 | |
222 | teardown $dir || return 1 | |
223 | ! test -d $dir || return 1 | |
224 | } | |
225 | ||
226 | ####################################################################### | |
227 | ||
228 | ## | |
229 | # Sends a signal to a single daemon. | |
230 | # This is a helper function for kill_daemons | |
231 | # | |
232 | # After the daemon is sent **signal**, its actual termination | |
233 | # will be verified by sending it signal 0. If the daemon is | |
234 | # still alive, kill_daemon will pause for a few seconds and | |
235 | # try again. This will repeat for a fixed number of times | |
236 | # before kill_daemon returns on failure. The list of | |
237 | # sleep intervals can be specified as **delays** and defaults | |
238 | # to: | |
239 | # | |
240 | # 0.1 0.2 1 1 1 2 3 5 5 5 10 10 20 60 60 60 120 | |
241 | # | |
242 | # This sequence is designed to run first a very short sleep time (0.1) | |
243 | # if the machine is fast enough and the daemon terminates in a fraction of a | |
244 | # second. The increasing sleep numbers should give plenty of time for | |
245 | # the daemon to die even on the slowest running machine. If a daemon | |
246 | # takes more than a few minutes to stop (the sum of all sleep times), | |
247 | # there probably is no point in waiting more and a number of things | |
248 | # are likely to go wrong anyway: better give up and return on error. | |
249 | # | |
250 | # @param pid the process id to send a signal | |
251 | # @param send_signal the signal to send | |
252 | # @param delays sequence of sleep times before failure | |
253 | # | |
254 | function kill_daemon() { | |
7c673cae FG |
255 | local pid=$(cat $1) |
256 | local send_signal=$2 | |
257 | local delays=${3:-0.1 0.2 1 1 1 2 3 5 5 5 10 10 20 60 60 60 120} | |
258 | local exit_code=1 | |
11fdf7f2 TL |
259 | # In order to try after the last large sleep add 0 at the end so we check |
260 | # one last time before dropping out of the loop | |
261 | for try in $delays 0 ; do | |
7c673cae FG |
262 | if kill -$send_signal $pid 2> /dev/null ; then |
263 | exit_code=1 | |
264 | else | |
265 | exit_code=0 | |
266 | break | |
267 | fi | |
268 | send_signal=0 | |
269 | sleep $try | |
270 | done; | |
271 | return $exit_code | |
272 | } | |
273 | ||
274 | function test_kill_daemon() { | |
275 | local dir=$1 | |
276 | setup $dir || return 1 | |
277 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
278 | run_mgr $dir x || return 1 | |
279 | run_osd $dir 0 || return 1 | |
280 | ||
281 | name_prefix=osd | |
282 | for pidfile in $(find $dir 2>/dev/null | grep $name_prefix'[^/]*\.pid') ; do | |
283 | # | |
284 | # sending signal 0 won't kill the daemon | |
285 | # waiting just for one second instead of the default schedule | |
286 | # allows us to quickly verify what happens when kill fails | |
287 | # to stop the daemon (i.e. it must return false) | |
288 | # | |
289 | ! kill_daemon $pidfile 0 1 || return 1 | |
290 | # | |
291 | # killing just the osd and verify the mon still is responsive | |
292 | # | |
293 | kill_daemon $pidfile TERM || return 1 | |
294 | done | |
295 | ||
7c673cae FG |
296 | name_prefix=mgr |
297 | for pidfile in $(find $dir 2>/dev/null | grep $name_prefix'[^/]*\.pid') ; do | |
298 | # | |
299 | # kill the mgr | |
300 | # | |
301 | kill_daemon $pidfile TERM || return 1 | |
302 | done | |
303 | ||
304 | name_prefix=mon | |
305 | for pidfile in $(find $dir 2>/dev/null | grep $name_prefix'[^/]*\.pid') ; do | |
306 | # | |
307 | # kill the mon and verify it cannot be reached | |
308 | # | |
309 | kill_daemon $pidfile TERM || return 1 | |
224ce89b | 310 | ! timeout 5 ceph status || return 1 |
7c673cae FG |
311 | done |
312 | ||
313 | teardown $dir || return 1 | |
314 | } | |
315 | ||
316 | ## | |
317 | # Kill all daemons for which a .pid file exists in **dir**. Each | |
318 | # daemon is sent a **signal** and kill_daemons waits for it to exit | |
319 | # during a few minutes. By default all daemons are killed. If a | |
320 | # **name_prefix** is provided, only the daemons for which a pid | |
321 | # file is found matching the prefix are killed. See run_osd and | |
322 | # run_mon for more information about the name conventions for | |
323 | # the pid files. | |
324 | # | |
325 | # Send TERM to all daemons : kill_daemons $dir | |
326 | # Send KILL to all daemons : kill_daemons $dir KILL | |
327 | # Send KILL to all osds : kill_daemons $dir KILL osd | |
328 | # Send KILL to osd 1 : kill_daemons $dir KILL osd.1 | |
329 | # | |
330 | # If a daemon is sent the TERM signal and does not terminate | |
331 | # within a few minutes, it will still be running even after | |
c07f9fc5 | 332 | # kill_daemons returns. |
7c673cae FG |
333 | # |
334 | # If all daemons are kill successfully the function returns 0 | |
c07f9fc5 | 335 | # if at least one daemon remains, this is treated as an |
7c673cae FG |
336 | # error and the function return 1. |
337 | # | |
338 | # @param dir path name of the environment | |
339 | # @param signal name of the first signal (defaults to TERM) | |
340 | # @param name_prefix only kill match daemons (defaults to all) | |
341 | # @param delays sequence of sleep times before failure | |
342 | # @return 0 on success, 1 on error | |
343 | # | |
344 | function kill_daemons() { | |
345 | local trace=$(shopt -q -o xtrace && echo true || echo false) | |
346 | $trace && shopt -u -o xtrace | |
347 | local dir=$1 | |
348 | local signal=${2:-TERM} | |
349 | local name_prefix=$3 # optional, osd, mon, osd.1 | |
350 | local delays=$4 #optional timing | |
351 | local status=0 | |
352 | local pids="" | |
353 | ||
354 | for pidfile in $(find $dir 2>/dev/null | grep $name_prefix'[^/]*\.pid') ; do | |
355 | run_in_background pids kill_daemon $pidfile $signal $delays | |
356 | done | |
357 | ||
358 | wait_background pids | |
359 | status=$? | |
360 | ||
361 | $trace && shopt -s -o xtrace | |
362 | return $status | |
363 | } | |
364 | ||
365 | function test_kill_daemons() { | |
366 | local dir=$1 | |
367 | setup $dir || return 1 | |
368 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
369 | run_mgr $dir x || return 1 | |
370 | run_osd $dir 0 || return 1 | |
371 | # | |
372 | # sending signal 0 won't kill the daemon | |
373 | # waiting just for one second instead of the default schedule | |
c07f9fc5 | 374 | # allows us to quickly verify what happens when kill fails |
7c673cae FG |
375 | # to stop the daemon (i.e. it must return false) |
376 | # | |
377 | ! kill_daemons $dir 0 osd 1 || return 1 | |
378 | # | |
379 | # killing just the osd and verify the mon still is responsive | |
380 | # | |
381 | kill_daemons $dir TERM osd || return 1 | |
7c673cae FG |
382 | # |
383 | # kill the mgr | |
384 | # | |
385 | kill_daemons $dir TERM mgr || return 1 | |
386 | # | |
387 | # kill the mon and verify it cannot be reached | |
388 | # | |
389 | kill_daemons $dir TERM || return 1 | |
224ce89b | 390 | ! timeout 5 ceph status || return 1 |
7c673cae FG |
391 | teardown $dir || return 1 |
392 | } | |
393 | ||
a8e16298 TL |
394 | # |
395 | # return a random TCP port which is not used yet | |
396 | # | |
397 | # please note, there could be racing if we use this function for | |
398 | # a free port, and then try to bind on this port. | |
399 | # | |
400 | function get_unused_port() { | |
401 | local ip=127.0.0.1 | |
402 | python3 -c "import socket; s=socket.socket(); s.bind(('$ip', 0)); print(s.getsockname()[1]); s.close()" | |
403 | } | |
404 | ||
7c673cae FG |
405 | ####################################################################### |
406 | ||
407 | ## | |
408 | # Run a monitor by the name mon.**id** with data in **dir**/**id**. | |
409 | # The logs can be found in **dir**/mon.**id**.log and the pid file | |
410 | # is **dir**/mon.**id**.pid and the admin socket is | |
411 | # **dir**/**id**/ceph-mon.**id**.asok. | |
412 | # | |
413 | # The remaining arguments are passed verbatim to ceph-mon --mkfs | |
414 | # and the ceph-mon daemon. | |
415 | # | |
416 | # Two mandatory arguments must be provided: --fsid and --mon-host | |
417 | # Instead of adding them to every call to run_mon, they can be | |
418 | # set in the CEPH_ARGS environment variable to be read implicitly | |
419 | # by every ceph command. | |
420 | # | |
421 | # The CEPH_CONF variable is expected to be set to /dev/null to | |
422 | # only rely on arguments for configuration. | |
423 | # | |
424 | # Examples: | |
425 | # | |
426 | # CEPH_ARGS="--fsid=$(uuidgen) " | |
427 | # CEPH_ARGS+="--mon-host=127.0.0.1:7018 " | |
428 | # run_mon $dir a # spawn a mon and bind port 7018 | |
429 | # run_mon $dir a --debug-filestore=20 # spawn with filestore debugging | |
430 | # | |
431 | # If mon_initial_members is not set, the default rbd pool is deleted | |
432 | # and replaced with a replicated pool with less placement groups to | |
433 | # speed up initialization. If mon_initial_members is set, no attempt | |
434 | # is made to recreate the rbd pool because it would hang forever, | |
435 | # waiting for other mons to join. | |
436 | # | |
437 | # A **dir**/ceph.conf file is created but not meant to be used by any | |
438 | # function. It is convenient for debugging a failure with: | |
439 | # | |
440 | # ceph --conf **dir**/ceph.conf -s | |
441 | # | |
442 | # @param dir path name of the environment | |
443 | # @param id mon identifier | |
444 | # @param ... can be any option valid for ceph-mon | |
445 | # @return 0 on success, 1 on error | |
446 | # | |
c07f9fc5 | 447 | function run_mon() { |
7c673cae FG |
448 | local dir=$1 |
449 | shift | |
450 | local id=$1 | |
451 | shift | |
452 | local data=$dir/$id | |
453 | ||
454 | ceph-mon \ | |
455 | --id $id \ | |
456 | --mkfs \ | |
457 | --mon-data=$data \ | |
458 | --run-dir=$dir \ | |
459 | "$@" || return 1 | |
460 | ||
461 | ceph-mon \ | |
462 | --id $id \ | |
11fdf7f2 | 463 | --osd-failsafe-full-ratio=.99 \ |
7c673cae FG |
464 | --mon-osd-full-ratio=.99 \ |
465 | --mon-data-avail-crit=1 \ | |
b5b8bbf5 | 466 | --mon-data-avail-warn=5 \ |
7c673cae FG |
467 | --paxos-propose-interval=0.1 \ |
468 | --osd-crush-chooseleaf-type=0 \ | |
c07f9fc5 | 469 | $EXTRA_OPTS \ |
7c673cae FG |
470 | --debug-mon 20 \ |
471 | --debug-ms 20 \ | |
472 | --debug-paxos 20 \ | |
473 | --chdir= \ | |
474 | --mon-data=$data \ | |
475 | --log-file=$dir/\$name.log \ | |
c07f9fc5 | 476 | --admin-socket=$(get_asok_path) \ |
7c673cae FG |
477 | --mon-cluster-log-file=$dir/log \ |
478 | --run-dir=$dir \ | |
479 | --pid-file=$dir/\$name.pid \ | |
480 | --mon-allow-pool-delete \ | |
c07f9fc5 | 481 | --mon-osd-backfillfull-ratio .99 \ |
7c673cae FG |
482 | "$@" || return 1 |
483 | ||
484 | cat > $dir/ceph.conf <<EOF | |
485 | [global] | |
486 | fsid = $(get_config mon $id fsid) | |
487 | mon host = $(get_config mon $id mon_host) | |
488 | EOF | |
224ce89b WB |
489 | } |
490 | ||
7c673cae FG |
491 | function test_run_mon() { |
492 | local dir=$1 | |
493 | ||
494 | setup $dir || return 1 | |
495 | ||
496 | run_mon $dir a --mon-initial-members=a || return 1 | |
11fdf7f2 | 497 | ceph mon dump | grep "mon.a" || return 1 |
7c673cae FG |
498 | kill_daemons $dir || return 1 |
499 | ||
11fdf7f2 TL |
500 | run_mon $dir a --osd_pool_default_size=3 || return 1 |
501 | run_osd $dir 0 || return 1 | |
502 | run_osd $dir 1 || return 1 | |
503 | run_osd $dir 2 || return 1 | |
c07f9fc5 | 504 | create_rbd_pool || return 1 |
11fdf7f2 | 505 | ceph osd dump | grep "pool 1 'rbd'" || return 1 |
c07f9fc5 | 506 | local size=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path mon.a) \ |
7c673cae FG |
507 | config get osd_pool_default_size) |
508 | test "$size" = '{"osd_pool_default_size":"3"}' || return 1 | |
509 | ||
510 | ! CEPH_ARGS='' ceph status || return 1 | |
511 | CEPH_ARGS='' ceph --conf $dir/ceph.conf status || return 1 | |
512 | ||
513 | kill_daemons $dir || return 1 | |
514 | ||
515 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
c07f9fc5 | 516 | local size=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path mon.a) \ |
7c673cae FG |
517 | config get osd_pool_default_size) |
518 | test "$size" = '{"osd_pool_default_size":"1"}' || return 1 | |
519 | kill_daemons $dir || return 1 | |
520 | ||
521 | CEPH_ARGS="$CEPH_ARGS --osd_pool_default_size=2" \ | |
522 | run_mon $dir a || return 1 | |
c07f9fc5 | 523 | local size=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path mon.a) \ |
7c673cae FG |
524 | config get osd_pool_default_size) |
525 | test "$size" = '{"osd_pool_default_size":"2"}' || return 1 | |
526 | kill_daemons $dir || return 1 | |
527 | ||
528 | teardown $dir || return 1 | |
529 | } | |
530 | ||
c07f9fc5 FG |
531 | function create_rbd_pool() { |
532 | ceph osd pool delete rbd rbd --yes-i-really-really-mean-it || return 1 | |
b5b8bbf5 | 533 | create_pool rbd $PG_NUM || return 1 |
c07f9fc5 FG |
534 | rbd pool init rbd |
535 | } | |
536 | ||
b5b8bbf5 FG |
537 | function create_pool() { |
538 | ceph osd pool create "$@" | |
539 | sleep 1 | |
540 | } | |
541 | ||
28e407b8 AA |
542 | function delete_pool() { |
543 | local poolname=$1 | |
544 | ceph osd pool delete $poolname $poolname --yes-i-really-really-mean-it | |
545 | } | |
546 | ||
7c673cae FG |
547 | ####################################################################### |
548 | ||
549 | function run_mgr() { | |
550 | local dir=$1 | |
551 | shift | |
552 | local id=$1 | |
553 | shift | |
554 | local data=$dir/$id | |
555 | ||
556 | ceph-mgr \ | |
557 | --id $id \ | |
c07f9fc5 | 558 | $EXTRA_OPTS \ |
11fdf7f2 | 559 | --osd-failsafe-full-ratio=.99 \ |
7c673cae FG |
560 | --debug-mgr 20 \ |
561 | --debug-objecter 20 \ | |
562 | --debug-ms 20 \ | |
563 | --debug-paxos 20 \ | |
564 | --chdir= \ | |
565 | --mgr-data=$data \ | |
566 | --log-file=$dir/\$name.log \ | |
c07f9fc5 | 567 | --admin-socket=$(get_asok_path) \ |
7c673cae FG |
568 | --run-dir=$dir \ |
569 | --pid-file=$dir/\$name.pid \ | |
11fdf7f2 TL |
570 | --mgr-module-path=$(realpath ${CEPH_ROOT}/src/pybind/mgr) \ |
571 | "$@" || return 1 | |
572 | } | |
573 | ||
574 | function run_mds() { | |
575 | local dir=$1 | |
576 | shift | |
577 | local id=$1 | |
578 | shift | |
579 | local data=$dir/$id | |
580 | ||
581 | ceph-mds \ | |
582 | --id $id \ | |
583 | $EXTRA_OPTS \ | |
584 | --debug-mds 20 \ | |
585 | --debug-objecter 20 \ | |
586 | --debug-ms 20 \ | |
587 | --chdir= \ | |
588 | --mds-data=$data \ | |
589 | --log-file=$dir/\$name.log \ | |
590 | --admin-socket=$(get_asok_path) \ | |
591 | --run-dir=$dir \ | |
592 | --pid-file=$dir/\$name.pid \ | |
7c673cae FG |
593 | "$@" || return 1 |
594 | } | |
595 | ||
596 | ####################################################################### | |
597 | ||
598 | ## | |
599 | # Create (prepare) and run (activate) an osd by the name osd.**id** | |
600 | # with data in **dir**/**id**. The logs can be found in | |
601 | # **dir**/osd.**id**.log, the pid file is **dir**/osd.**id**.pid and | |
602 | # the admin socket is **dir**/**id**/ceph-osd.**id**.asok. | |
603 | # | |
604 | # The remaining arguments are passed verbatim to ceph-osd. | |
605 | # | |
606 | # Two mandatory arguments must be provided: --fsid and --mon-host | |
607 | # Instead of adding them to every call to run_osd, they can be | |
608 | # set in the CEPH_ARGS environment variable to be read implicitly | |
609 | # by every ceph command. | |
610 | # | |
611 | # The CEPH_CONF variable is expected to be set to /dev/null to | |
612 | # only rely on arguments for configuration. | |
613 | # | |
11fdf7f2 TL |
614 | # The run_osd function creates the OSD data directory on the **dir**/**id** |
615 | # directory and relies on the activate_osd function to run the daemon. | |
7c673cae FG |
616 | # |
617 | # Examples: | |
618 | # | |
619 | # CEPH_ARGS="--fsid=$(uuidgen) " | |
620 | # CEPH_ARGS+="--mon-host=127.0.0.1:7018 " | |
621 | # run_osd $dir 0 # prepare and activate an osd using the monitor listening on 7018 | |
622 | # | |
623 | # @param dir path name of the environment | |
624 | # @param id osd identifier | |
625 | # @param ... can be any option valid for ceph-osd | |
626 | # @return 0 on success, 1 on error | |
627 | # | |
628 | function run_osd() { | |
629 | local dir=$1 | |
630 | shift | |
631 | local id=$1 | |
632 | shift | |
633 | local osd_data=$dir/$id | |
634 | ||
11fdf7f2 TL |
635 | local ceph_args="$CEPH_ARGS" |
636 | ceph_args+=" --osd-failsafe-full-ratio=.99" | |
637 | ceph_args+=" --osd-journal-size=100" | |
638 | ceph_args+=" --osd-scrub-load-threshold=2000" | |
639 | ceph_args+=" --osd-data=$osd_data" | |
640 | ceph_args+=" --osd-journal=${osd_data}/journal" | |
641 | ceph_args+=" --chdir=" | |
642 | ceph_args+=$EXTRA_OPTS | |
643 | ceph_args+=" --run-dir=$dir" | |
644 | ceph_args+=" --admin-socket=$(get_asok_path)" | |
645 | ceph_args+=" --debug-osd=20" | |
646 | ceph_args+=" --log-file=$dir/\$name.log" | |
647 | ceph_args+=" --pid-file=$dir/\$name.pid" | |
648 | ceph_args+=" --osd-max-object-name-len=460" | |
649 | ceph_args+=" --osd-max-object-namespace-len=64" | |
650 | ceph_args+=" --enable-experimental-unrecoverable-data-corrupting-features=*" | |
651 | ceph_args+=" " | |
652 | ceph_args+="$@" | |
7c673cae | 653 | mkdir -p $osd_data |
7c673cae | 654 | |
11fdf7f2 TL |
655 | local uuid=`uuidgen` |
656 | echo "add osd$id $uuid" | |
657 | OSD_SECRET=$(ceph-authtool --gen-print-key) | |
658 | echo "{\"cephx_secret\": \"$OSD_SECRET\"}" > $osd_data/new.json | |
659 | ceph osd new $uuid -i $osd_data/new.json | |
660 | rm $osd_data/new.json | |
661 | ceph-osd -i $id $ceph_args --mkfs --key $OSD_SECRET --osd-uuid $uuid | |
662 | ||
663 | local key_fn=$osd_data/keyring | |
664 | cat > $key_fn<<EOF | |
665 | [osd.$id] | |
666 | key = $OSD_SECRET | |
667 | EOF | |
668 | echo adding osd$id key to auth repository | |
669 | ceph -i "$key_fn" auth add osd.$id osd "allow *" mon "allow profile osd" mgr "allow profile osd" | |
670 | echo start osd.$id | |
671 | ceph-osd -i $id $ceph_args & | |
672 | ||
eafe8130 TL |
673 | # If noup is set, then can't wait for this osd |
674 | if ceph osd dump --format=json | jq '.flags_set[]' | grep -q '"noup"' ; then | |
675 | return 0 | |
676 | fi | |
11fdf7f2 TL |
677 | wait_for_osd up $id || return 1 |
678 | ||
7c673cae FG |
679 | } |
680 | ||
eafe8130 | 681 | function run_osd_filestore() { |
7c673cae FG |
682 | local dir=$1 |
683 | shift | |
684 | local id=$1 | |
685 | shift | |
686 | local osd_data=$dir/$id | |
687 | ||
11fdf7f2 TL |
688 | local ceph_args="$CEPH_ARGS" |
689 | ceph_args+=" --osd-failsafe-full-ratio=.99" | |
690 | ceph_args+=" --osd-journal-size=100" | |
691 | ceph_args+=" --osd-scrub-load-threshold=2000" | |
692 | ceph_args+=" --osd-data=$osd_data" | |
693 | ceph_args+=" --osd-journal=${osd_data}/journal" | |
694 | ceph_args+=" --chdir=" | |
695 | ceph_args+=$EXTRA_OPTS | |
696 | ceph_args+=" --run-dir=$dir" | |
697 | ceph_args+=" --admin-socket=$(get_asok_path)" | |
698 | ceph_args+=" --debug-osd=20" | |
699 | ceph_args+=" --log-file=$dir/\$name.log" | |
700 | ceph_args+=" --pid-file=$dir/\$name.pid" | |
701 | ceph_args+=" --osd-max-object-name-len=460" | |
702 | ceph_args+=" --osd-max-object-namespace-len=64" | |
703 | ceph_args+=" --enable-experimental-unrecoverable-data-corrupting-features=*" | |
704 | ceph_args+=" " | |
705 | ceph_args+="$@" | |
7c673cae | 706 | mkdir -p $osd_data |
7c673cae | 707 | |
11fdf7f2 TL |
708 | local uuid=`uuidgen` |
709 | echo "add osd$osd $uuid" | |
710 | OSD_SECRET=$(ceph-authtool --gen-print-key) | |
711 | echo "{\"cephx_secret\": \"$OSD_SECRET\"}" > $osd_data/new.json | |
712 | ceph osd new $uuid -i $osd_data/new.json | |
713 | rm $osd_data/new.json | |
eafe8130 | 714 | ceph-osd -i $id $ceph_args --mkfs --key $OSD_SECRET --osd-uuid $uuid --osd-objectstore=filestore |
11fdf7f2 TL |
715 | |
716 | local key_fn=$osd_data/keyring | |
717 | cat > $key_fn<<EOF | |
718 | [osd.$osd] | |
719 | key = $OSD_SECRET | |
720 | EOF | |
721 | echo adding osd$id key to auth repository | |
722 | ceph -i "$key_fn" auth add osd.$id osd "allow *" mon "allow profile osd" mgr "allow profile osd" | |
723 | echo start osd.$id | |
724 | ceph-osd -i $id $ceph_args & | |
725 | ||
eafe8130 TL |
726 | # If noup is set, then can't wait for this osd |
727 | if ceph osd dump --format=json | jq '.flags_set[]' | grep -q '"noup"' ; then | |
728 | return 0 | |
729 | fi | |
11fdf7f2 TL |
730 | wait_for_osd up $id || return 1 |
731 | ||
732 | ||
7c673cae FG |
733 | } |
734 | ||
735 | function test_run_osd() { | |
736 | local dir=$1 | |
737 | ||
738 | setup $dir || return 1 | |
739 | ||
740 | run_mon $dir a || return 1 | |
741 | run_mgr $dir x || return 1 | |
742 | ||
743 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 744 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path osd.0) \ |
7c673cae FG |
745 | config get osd_max_backfills) |
746 | echo "$backfills" | grep --quiet 'osd_max_backfills' || return 1 | |
747 | ||
748 | run_osd $dir 1 --osd-max-backfills 20 || return 1 | |
c07f9fc5 | 749 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path osd.1) \ |
7c673cae FG |
750 | config get osd_max_backfills) |
751 | test "$backfills" = '{"osd_max_backfills":"20"}' || return 1 | |
752 | ||
753 | CEPH_ARGS="$CEPH_ARGS --osd-max-backfills 30" run_osd $dir 2 || return 1 | |
c07f9fc5 | 754 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path osd.2) \ |
7c673cae FG |
755 | config get osd_max_backfills) |
756 | test "$backfills" = '{"osd_max_backfills":"30"}' || return 1 | |
757 | ||
758 | teardown $dir || return 1 | |
759 | } | |
760 | ||
761 | ####################################################################### | |
762 | ||
763 | ## | |
764 | # Shutdown and remove all traces of the osd by the name osd.**id**. | |
765 | # | |
766 | # The OSD is shutdown with the TERM signal. It is then removed from | |
767 | # the auth list, crush map, osd map etc and the files associated with | |
768 | # it are also removed. | |
769 | # | |
770 | # @param dir path name of the environment | |
771 | # @param id osd identifier | |
772 | # @return 0 on success, 1 on error | |
773 | # | |
774 | function destroy_osd() { | |
775 | local dir=$1 | |
776 | local id=$2 | |
777 | ||
7c673cae | 778 | ceph osd out osd.$id || return 1 |
c07f9fc5 | 779 | kill_daemons $dir TERM osd.$id || return 1 |
92f5a8d4 | 780 | ceph osd down osd.$id || return 1 |
c07f9fc5 | 781 | ceph osd purge osd.$id --yes-i-really-mean-it || return 1 |
7c673cae FG |
782 | teardown $dir/$id || return 1 |
783 | rm -fr $dir/$id | |
784 | } | |
785 | ||
786 | function test_destroy_osd() { | |
787 | local dir=$1 | |
788 | ||
789 | setup $dir || return 1 | |
790 | run_mon $dir a || return 1 | |
791 | run_mgr $dir x || return 1 | |
792 | run_osd $dir 0 || return 1 | |
793 | destroy_osd $dir 0 || return 1 | |
794 | ! ceph osd dump | grep "osd.$id " || return 1 | |
795 | teardown $dir || return 1 | |
796 | } | |
797 | ||
798 | ####################################################################### | |
799 | ||
800 | ## | |
801 | # Run (activate) an osd by the name osd.**id** with data in | |
802 | # **dir**/**id**. The logs can be found in **dir**/osd.**id**.log, | |
803 | # the pid file is **dir**/osd.**id**.pid and the admin socket is | |
804 | # **dir**/**id**/ceph-osd.**id**.asok. | |
805 | # | |
806 | # The remaining arguments are passed verbatim to ceph-osd. | |
807 | # | |
808 | # Two mandatory arguments must be provided: --fsid and --mon-host | |
809 | # Instead of adding them to every call to activate_osd, they can be | |
810 | # set in the CEPH_ARGS environment variable to be read implicitly | |
811 | # by every ceph command. | |
812 | # | |
813 | # The CEPH_CONF variable is expected to be set to /dev/null to | |
814 | # only rely on arguments for configuration. | |
815 | # | |
816 | # The activate_osd function expects a valid OSD data directory | |
817 | # in **dir**/**id**, either just created via run_osd or re-using | |
818 | # one left by a previous run of ceph-osd. The ceph-osd daemon is | |
11fdf7f2 | 819 | # run directly on the foreground |
7c673cae FG |
820 | # |
821 | # The activate_osd function blocks until the monitor reports the osd | |
822 | # up. If it fails to do so within $TIMEOUT seconds, activate_osd | |
823 | # fails. | |
824 | # | |
825 | # Examples: | |
826 | # | |
827 | # CEPH_ARGS="--fsid=$(uuidgen) " | |
828 | # CEPH_ARGS+="--mon-host=127.0.0.1:7018 " | |
829 | # activate_osd $dir 0 # activate an osd using the monitor listening on 7018 | |
830 | # | |
831 | # @param dir path name of the environment | |
832 | # @param id osd identifier | |
833 | # @param ... can be any option valid for ceph-osd | |
834 | # @return 0 on success, 1 on error | |
835 | # | |
836 | function activate_osd() { | |
837 | local dir=$1 | |
838 | shift | |
839 | local id=$1 | |
840 | shift | |
841 | local osd_data=$dir/$id | |
842 | ||
7c673cae | 843 | local ceph_args="$CEPH_ARGS" |
7c673cae FG |
844 | ceph_args+=" --osd-failsafe-full-ratio=.99" |
845 | ceph_args+=" --osd-journal-size=100" | |
846 | ceph_args+=" --osd-scrub-load-threshold=2000" | |
847 | ceph_args+=" --osd-data=$osd_data" | |
11fdf7f2 | 848 | ceph_args+=" --osd-journal=${osd_data}/journal" |
7c673cae | 849 | ceph_args+=" --chdir=" |
c07f9fc5 | 850 | ceph_args+=$EXTRA_OPTS |
7c673cae | 851 | ceph_args+=" --run-dir=$dir" |
c07f9fc5 | 852 | ceph_args+=" --admin-socket=$(get_asok_path)" |
7c673cae FG |
853 | ceph_args+=" --debug-osd=20" |
854 | ceph_args+=" --log-file=$dir/\$name.log" | |
855 | ceph_args+=" --pid-file=$dir/\$name.pid" | |
11fdf7f2 TL |
856 | ceph_args+=" --osd-max-object-name-len=460" |
857 | ceph_args+=" --osd-max-object-namespace-len=64" | |
858 | ceph_args+=" --enable-experimental-unrecoverable-data-corrupting-features=*" | |
7c673cae FG |
859 | ceph_args+=" " |
860 | ceph_args+="$@" | |
861 | mkdir -p $osd_data | |
11fdf7f2 TL |
862 | |
863 | echo start osd.$id | |
864 | ceph-osd -i $id $ceph_args & | |
7c673cae FG |
865 | |
866 | [ "$id" = "$(cat $osd_data/whoami)" ] || return 1 | |
867 | ||
eafe8130 TL |
868 | # If noup is set, then can't wait for this osd |
869 | if ceph osd dump --format=json | jq '.flags_set[]' | grep -q '"noup"' ; then | |
870 | return 0 | |
871 | fi | |
7c673cae FG |
872 | wait_for_osd up $id || return 1 |
873 | } | |
874 | ||
875 | function test_activate_osd() { | |
876 | local dir=$1 | |
877 | ||
878 | setup $dir || return 1 | |
879 | ||
880 | run_mon $dir a || return 1 | |
881 | run_mgr $dir x || return 1 | |
882 | ||
883 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 884 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path osd.0) \ |
7c673cae FG |
885 | config get osd_max_backfills) |
886 | echo "$backfills" | grep --quiet 'osd_max_backfills' || return 1 | |
887 | ||
888 | kill_daemons $dir TERM osd || return 1 | |
889 | ||
890 | activate_osd $dir 0 --osd-max-backfills 20 || return 1 | |
c07f9fc5 | 891 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path osd.0) \ |
7c673cae FG |
892 | config get osd_max_backfills) |
893 | test "$backfills" = '{"osd_max_backfills":"20"}' || return 1 | |
894 | ||
895 | teardown $dir || return 1 | |
896 | } | |
897 | ||
898 | ####################################################################### | |
899 | ||
900 | ## | |
901 | # Wait until the OSD **id** is either up or down, as specified by | |
902 | # **state**. It fails after $TIMEOUT seconds. | |
903 | # | |
904 | # @param state either up or down | |
905 | # @param id osd identifier | |
906 | # @return 0 on success, 1 on error | |
907 | # | |
908 | function wait_for_osd() { | |
909 | local state=$1 | |
910 | local id=$2 | |
911 | ||
912 | status=1 | |
913 | for ((i=0; i < $TIMEOUT; i++)); do | |
914 | echo $i | |
915 | if ! ceph osd dump | grep "osd.$id $state"; then | |
916 | sleep 1 | |
917 | else | |
918 | status=0 | |
919 | break | |
920 | fi | |
921 | done | |
922 | return $status | |
923 | } | |
924 | ||
925 | function test_wait_for_osd() { | |
926 | local dir=$1 | |
927 | setup $dir || return 1 | |
928 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
929 | run_mgr $dir x || return 1 | |
930 | run_osd $dir 0 || return 1 | |
92f5a8d4 | 931 | run_osd $dir 1 || return 1 |
7c673cae | 932 | wait_for_osd up 0 || return 1 |
92f5a8d4 TL |
933 | wait_for_osd up 1 || return 1 |
934 | kill_daemons $dir TERM osd.0 || return 1 | |
7c673cae FG |
935 | wait_for_osd down 0 || return 1 |
936 | ( TIMEOUT=1 ; ! wait_for_osd up 0 ) || return 1 | |
937 | teardown $dir || return 1 | |
938 | } | |
939 | ||
940 | ####################################################################### | |
941 | ||
942 | ## | |
943 | # Display the list of OSD ids supporting the **objectname** stored in | |
944 | # **poolname**, as reported by ceph osd map. | |
945 | # | |
946 | # @param poolname an existing pool | |
947 | # @param objectname an objectname (may or may not exist) | |
948 | # @param STDOUT white space separated list of OSD ids | |
949 | # @return 0 on success, 1 on error | |
950 | # | |
951 | function get_osds() { | |
952 | local poolname=$1 | |
953 | local objectname=$2 | |
954 | ||
31f18b77 FG |
955 | local osds=$(ceph --format json osd map $poolname $objectname 2>/dev/null | \ |
956 | jq '.acting | .[]') | |
7c673cae FG |
957 | # get rid of the trailing space |
958 | echo $osds | |
959 | } | |
960 | ||
961 | function test_get_osds() { | |
962 | local dir=$1 | |
963 | ||
964 | setup $dir || return 1 | |
965 | run_mon $dir a --osd_pool_default_size=2 || return 1 | |
966 | run_mgr $dir x || return 1 | |
967 | run_osd $dir 0 || return 1 | |
968 | run_osd $dir 1 || return 1 | |
c07f9fc5 | 969 | create_rbd_pool || return 1 |
7c673cae | 970 | wait_for_clean || return 1 |
c07f9fc5 | 971 | create_rbd_pool || return 1 |
7c673cae FG |
972 | get_osds rbd GROUP | grep --quiet '^[0-1] [0-1]$' || return 1 |
973 | teardown $dir || return 1 | |
974 | } | |
975 | ||
976 | ####################################################################### | |
977 | ||
978 | ## | |
979 | # Wait for the monitor to form quorum (optionally, of size N) | |
980 | # | |
981 | # @param timeout duration (lower-bound) to wait for quorum to be formed | |
982 | # @param quorumsize size of quorum to wait for | |
983 | # @return 0 on success, 1 on error | |
984 | # | |
985 | function wait_for_quorum() { | |
986 | local timeout=$1 | |
987 | local quorumsize=$2 | |
988 | ||
989 | if [[ -z "$timeout" ]]; then | |
990 | timeout=300 | |
991 | fi | |
992 | ||
993 | if [[ -z "$quorumsize" ]]; then | |
994 | timeout $timeout ceph mon_status --format=json >&/dev/null || return 1 | |
995 | return 0 | |
996 | fi | |
997 | ||
998 | no_quorum=1 | |
c07f9fc5 | 999 | wait_until=$((`date +%s` + $timeout)) |
7c673cae FG |
1000 | while [[ $(date +%s) -lt $wait_until ]]; do |
1001 | jqfilter='.quorum | length == '$quorumsize | |
1002 | jqinput="$(timeout $timeout ceph mon_status --format=json 2>/dev/null)" | |
1003 | res=$(echo $jqinput | jq "$jqfilter") | |
1004 | if [[ "$res" == "true" ]]; then | |
1005 | no_quorum=0 | |
1006 | break | |
1007 | fi | |
1008 | done | |
1009 | return $no_quorum | |
1010 | } | |
1011 | ||
1012 | ####################################################################### | |
1013 | ||
1014 | ## | |
1015 | # Return the PG of supporting the **objectname** stored in | |
1016 | # **poolname**, as reported by ceph osd map. | |
1017 | # | |
1018 | # @param poolname an existing pool | |
1019 | # @param objectname an objectname (may or may not exist) | |
1020 | # @param STDOUT a PG | |
1021 | # @return 0 on success, 1 on error | |
1022 | # | |
1023 | function get_pg() { | |
1024 | local poolname=$1 | |
1025 | local objectname=$2 | |
1026 | ||
31f18b77 | 1027 | ceph --format json osd map $poolname $objectname 2>/dev/null | jq -r '.pgid' |
7c673cae FG |
1028 | } |
1029 | ||
1030 | function test_get_pg() { | |
1031 | local dir=$1 | |
1032 | ||
1033 | setup $dir || return 1 | |
1034 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1035 | run_mgr $dir x || return 1 | |
1036 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1037 | create_rbd_pool || return 1 |
7c673cae FG |
1038 | wait_for_clean || return 1 |
1039 | get_pg rbd GROUP | grep --quiet '^[0-9]\.[0-9a-f][0-9a-f]*$' || return 1 | |
1040 | teardown $dir || return 1 | |
1041 | } | |
1042 | ||
1043 | ####################################################################### | |
1044 | ||
1045 | ## | |
1046 | # Return the value of the **config**, obtained via the config get command | |
1047 | # of the admin socket of **daemon**.**id**. | |
1048 | # | |
1049 | # @param daemon mon or osd | |
1050 | # @param id mon or osd ID | |
1051 | # @param config the configuration variable name as found in config_opts.h | |
1052 | # @param STDOUT the config value | |
1053 | # @return 0 on success, 1 on error | |
1054 | # | |
1055 | function get_config() { | |
1056 | local daemon=$1 | |
1057 | local id=$2 | |
1058 | local config=$3 | |
1059 | ||
1060 | CEPH_ARGS='' \ | |
c07f9fc5 | 1061 | ceph --format json daemon $(get_asok_path $daemon.$id) \ |
7c673cae | 1062 | config get $config 2> /dev/null | \ |
31f18b77 | 1063 | jq -r ".$config" |
7c673cae FG |
1064 | } |
1065 | ||
1066 | function test_get_config() { | |
1067 | local dir=$1 | |
1068 | ||
1069 | # override the default config using command line arg and check it | |
1070 | setup $dir || return 1 | |
1071 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1072 | test $(get_config mon a osd_pool_default_size) = 1 || return 1 | |
1073 | run_mgr $dir x || return 1 | |
1074 | run_osd $dir 0 --osd_max_scrubs=3 || return 1 | |
1075 | test $(get_config osd 0 osd_max_scrubs) = 3 || return 1 | |
1076 | teardown $dir || return 1 | |
1077 | } | |
1078 | ||
1079 | ####################################################################### | |
1080 | ||
1081 | ## | |
1082 | # Set the **config** to specified **value**, via the config set command | |
1083 | # of the admin socket of **daemon**.**id** | |
1084 | # | |
1085 | # @param daemon mon or osd | |
1086 | # @param id mon or osd ID | |
1087 | # @param config the configuration variable name as found in config_opts.h | |
1088 | # @param value the config value | |
1089 | # @return 0 on success, 1 on error | |
1090 | # | |
1091 | function set_config() { | |
1092 | local daemon=$1 | |
1093 | local id=$2 | |
1094 | local config=$3 | |
1095 | local value=$4 | |
1096 | ||
c07f9fc5 | 1097 | test $(env CEPH_ARGS='' ceph --format json daemon $(get_asok_path $daemon.$id) \ |
31f18b77 FG |
1098 | config set $config $value 2> /dev/null | \ |
1099 | jq 'has("success")') == true | |
7c673cae FG |
1100 | } |
1101 | ||
1102 | function test_set_config() { | |
1103 | local dir=$1 | |
1104 | ||
1105 | setup $dir || return 1 | |
1106 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1107 | test $(get_config mon a ms_crc_header) = true || return 1 | |
1108 | set_config mon a ms_crc_header false || return 1 | |
1109 | test $(get_config mon a ms_crc_header) = false || return 1 | |
1110 | set_config mon a ms_crc_header true || return 1 | |
1111 | test $(get_config mon a ms_crc_header) = true || return 1 | |
1112 | teardown $dir || return 1 | |
1113 | } | |
1114 | ||
1115 | ####################################################################### | |
1116 | ||
1117 | ## | |
1118 | # Return the OSD id of the primary OSD supporting the **objectname** | |
1119 | # stored in **poolname**, as reported by ceph osd map. | |
1120 | # | |
1121 | # @param poolname an existing pool | |
1122 | # @param objectname an objectname (may or may not exist) | |
1123 | # @param STDOUT the primary OSD id | |
1124 | # @return 0 on success, 1 on error | |
1125 | # | |
1126 | function get_primary() { | |
1127 | local poolname=$1 | |
1128 | local objectname=$2 | |
1129 | ||
31f18b77 FG |
1130 | ceph --format json osd map $poolname $objectname 2>/dev/null | \ |
1131 | jq '.acting_primary' | |
7c673cae FG |
1132 | } |
1133 | ||
1134 | function test_get_primary() { | |
1135 | local dir=$1 | |
1136 | ||
1137 | setup $dir || return 1 | |
1138 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1139 | local osd=0 | |
1140 | run_mgr $dir x || return 1 | |
1141 | run_osd $dir $osd || return 1 | |
c07f9fc5 | 1142 | create_rbd_pool || return 1 |
7c673cae FG |
1143 | wait_for_clean || return 1 |
1144 | test $(get_primary rbd GROUP) = $osd || return 1 | |
1145 | teardown $dir || return 1 | |
1146 | } | |
1147 | ||
1148 | ####################################################################### | |
1149 | ||
1150 | ## | |
1151 | # Return the id of any OSD supporting the **objectname** stored in | |
1152 | # **poolname**, as reported by ceph osd map, except the primary. | |
1153 | # | |
1154 | # @param poolname an existing pool | |
1155 | # @param objectname an objectname (may or may not exist) | |
1156 | # @param STDOUT the OSD id | |
1157 | # @return 0 on success, 1 on error | |
1158 | # | |
1159 | function get_not_primary() { | |
1160 | local poolname=$1 | |
1161 | local objectname=$2 | |
1162 | ||
1163 | local primary=$(get_primary $poolname $objectname) | |
31f18b77 FG |
1164 | ceph --format json osd map $poolname $objectname 2>/dev/null | \ |
1165 | jq ".acting | map(select (. != $primary)) | .[0]" | |
7c673cae FG |
1166 | } |
1167 | ||
1168 | function test_get_not_primary() { | |
1169 | local dir=$1 | |
1170 | ||
1171 | setup $dir || return 1 | |
1172 | run_mon $dir a --osd_pool_default_size=2 || return 1 | |
1173 | run_mgr $dir x || return 1 | |
1174 | run_osd $dir 0 || return 1 | |
1175 | run_osd $dir 1 || return 1 | |
c07f9fc5 | 1176 | create_rbd_pool || return 1 |
7c673cae FG |
1177 | wait_for_clean || return 1 |
1178 | local primary=$(get_primary rbd GROUP) | |
1179 | local not_primary=$(get_not_primary rbd GROUP) | |
1180 | test $not_primary != $primary || return 1 | |
1181 | test $not_primary = 0 -o $not_primary = 1 || return 1 | |
1182 | teardown $dir || return 1 | |
1183 | } | |
1184 | ||
1185 | ####################################################################### | |
1186 | ||
11fdf7f2 TL |
1187 | function _objectstore_tool_nodown() { |
1188 | local dir=$1 | |
1189 | shift | |
1190 | local id=$1 | |
1191 | shift | |
1192 | local osd_data=$dir/$id | |
1193 | ||
11fdf7f2 TL |
1194 | ceph-objectstore-tool \ |
1195 | --data-path $osd_data \ | |
11fdf7f2 TL |
1196 | "$@" || return 1 |
1197 | } | |
1198 | ||
1199 | function _objectstore_tool_nowait() { | |
1200 | local dir=$1 | |
1201 | shift | |
1202 | local id=$1 | |
1203 | shift | |
1204 | ||
1205 | kill_daemons $dir TERM osd.$id >&2 < /dev/null || return 1 | |
1206 | ||
1207 | _objectstore_tool_nodown $dir $id "$@" || return 1 | |
1208 | activate_osd $dir $id $ceph_osd_args >&2 || return 1 | |
1209 | } | |
1210 | ||
7c673cae FG |
1211 | ## |
1212 | # Run ceph-objectstore-tool against the OSD **id** using the data path | |
1213 | # **dir**. The OSD is killed with TERM prior to running | |
1214 | # ceph-objectstore-tool because access to the data path is | |
1215 | # exclusive. The OSD is restarted after the command completes. The | |
1216 | # objectstore_tool returns after all PG are active+clean again. | |
1217 | # | |
1218 | # @param dir the data path of the OSD | |
1219 | # @param id the OSD id | |
1220 | # @param ... arguments to ceph-objectstore-tool | |
1221 | # @param STDIN the input of ceph-objectstore-tool | |
1222 | # @param STDOUT the output of ceph-objectstore-tool | |
1223 | # @return 0 on success, 1 on error | |
1224 | # | |
1225 | # The value of $ceph_osd_args will be passed to restarted osds | |
1226 | # | |
1227 | function objectstore_tool() { | |
1228 | local dir=$1 | |
1229 | shift | |
1230 | local id=$1 | |
1231 | shift | |
7c673cae | 1232 | |
11fdf7f2 | 1233 | _objectstore_tool_nowait $dir $id "$@" || return 1 |
7c673cae FG |
1234 | wait_for_clean >&2 |
1235 | } | |
1236 | ||
1237 | function test_objectstore_tool() { | |
1238 | local dir=$1 | |
1239 | ||
1240 | setup $dir || return 1 | |
1241 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1242 | local osd=0 | |
1243 | run_mgr $dir x || return 1 | |
1244 | run_osd $dir $osd || return 1 | |
c07f9fc5 | 1245 | create_rbd_pool || return 1 |
7c673cae FG |
1246 | wait_for_clean || return 1 |
1247 | rados --pool rbd put GROUP /etc/group || return 1 | |
1248 | objectstore_tool $dir $osd GROUP get-bytes | \ | |
1249 | diff - /etc/group | |
1250 | ! objectstore_tool $dir $osd NOTEXISTS get-bytes || return 1 | |
1251 | teardown $dir || return 1 | |
1252 | } | |
1253 | ||
1254 | ####################################################################### | |
1255 | ||
1256 | ## | |
1257 | # Predicate checking if there is an ongoing recovery in the | |
1258 | # cluster. If any of the recovering_{keys,bytes,objects}_per_sec | |
1259 | # counters are reported by ceph status, it means recovery is in | |
1260 | # progress. | |
1261 | # | |
1262 | # @return 0 if recovery in progress, 1 otherwise | |
1263 | # | |
1264 | function get_is_making_recovery_progress() { | |
31f18b77 FG |
1265 | local recovery_progress |
1266 | recovery_progress+=".recovering_keys_per_sec + " | |
1267 | recovery_progress+=".recovering_bytes_per_sec + " | |
1268 | recovery_progress+=".recovering_objects_per_sec" | |
1269 | local progress=$(ceph --format json status 2>/dev/null | \ | |
1270 | jq -r ".pgmap | $recovery_progress") | |
1271 | test "$progress" != null | |
7c673cae FG |
1272 | } |
1273 | ||
1274 | function test_get_is_making_recovery_progress() { | |
1275 | local dir=$1 | |
1276 | ||
1277 | setup $dir || return 1 | |
1278 | run_mon $dir a || return 1 | |
1279 | run_mgr $dir x || return 1 | |
1280 | ! get_is_making_recovery_progress || return 1 | |
1281 | teardown $dir || return 1 | |
1282 | } | |
1283 | ||
1284 | ####################################################################### | |
1285 | ||
1286 | ## | |
1287 | # Return the number of active PGs in the cluster. A PG is active if | |
1288 | # ceph pg dump pgs reports it both **active** and **clean** and that | |
1289 | # not **stale**. | |
1290 | # | |
1291 | # @param STDOUT the number of active PGs | |
1292 | # @return 0 on success, 1 on error | |
1293 | # | |
1294 | function get_num_active_clean() { | |
31f18b77 FG |
1295 | local expression |
1296 | expression+="select(contains(\"active\") and contains(\"clean\")) | " | |
1297 | expression+="select(contains(\"stale\") | not)" | |
1298 | ceph --format json pg dump pgs 2>/dev/null | \ | |
11fdf7f2 | 1299 | jq ".pg_stats | [.[] | .state | $expression] | length" |
7c673cae FG |
1300 | } |
1301 | ||
1302 | function test_get_num_active_clean() { | |
1303 | local dir=$1 | |
1304 | ||
1305 | setup $dir || return 1 | |
1306 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1307 | run_mgr $dir x || return 1 | |
1308 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1309 | create_rbd_pool || return 1 |
7c673cae FG |
1310 | wait_for_clean || return 1 |
1311 | local num_active_clean=$(get_num_active_clean) | |
1312 | test "$num_active_clean" = $PG_NUM || return 1 | |
1313 | teardown $dir || return 1 | |
1314 | } | |
1315 | ||
92f5a8d4 TL |
1316 | ## |
1317 | # Return the number of active or peered PGs in the cluster. A PG matches if | |
1318 | # ceph pg dump pgs reports it is either **active** or **peered** and that | |
1319 | # not **stale**. | |
1320 | # | |
1321 | # @param STDOUT the number of active PGs | |
1322 | # @return 0 on success, 1 on error | |
1323 | # | |
1324 | function get_num_active_or_peered() { | |
1325 | local expression | |
1326 | expression+="select(contains(\"active\") or contains(\"peered\")) | " | |
1327 | expression+="select(contains(\"stale\") | not)" | |
1328 | ceph --format json pg dump pgs 2>/dev/null | \ | |
1329 | jq ".pg_stats | [.[] | .state | $expression] | length" | |
1330 | } | |
1331 | ||
1332 | function test_get_num_active_or_peered() { | |
1333 | local dir=$1 | |
1334 | ||
1335 | setup $dir || return 1 | |
1336 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1337 | run_mgr $dir x || return 1 | |
1338 | run_osd $dir 0 || return 1 | |
1339 | create_rbd_pool || return 1 | |
1340 | wait_for_clean || return 1 | |
1341 | local num_peered=$(get_num_active_or_peered) | |
1342 | test "$num_peered" = $PG_NUM || return 1 | |
1343 | teardown $dir || return 1 | |
1344 | } | |
1345 | ||
7c673cae FG |
1346 | ####################################################################### |
1347 | ||
1348 | ## | |
1349 | # Return the number of PGs in the cluster, according to | |
1350 | # ceph pg dump pgs. | |
1351 | # | |
1352 | # @param STDOUT the number of PGs | |
1353 | # @return 0 on success, 1 on error | |
1354 | # | |
1355 | function get_num_pgs() { | |
31f18b77 | 1356 | ceph --format json status 2>/dev/null | jq '.pgmap.num_pgs' |
7c673cae FG |
1357 | } |
1358 | ||
1359 | function test_get_num_pgs() { | |
1360 | local dir=$1 | |
1361 | ||
1362 | setup $dir || return 1 | |
1363 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1364 | run_mgr $dir x || return 1 | |
1365 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1366 | create_rbd_pool || return 1 |
7c673cae FG |
1367 | wait_for_clean || return 1 |
1368 | local num_pgs=$(get_num_pgs) | |
1369 | test "$num_pgs" -gt 0 || return 1 | |
1370 | teardown $dir || return 1 | |
1371 | } | |
1372 | ||
1373 | ####################################################################### | |
1374 | ||
c07f9fc5 FG |
1375 | ## |
1376 | # Return the OSD ids in use by at least one PG in the cluster (either | |
1377 | # in the up or the acting set), according to ceph pg dump pgs. Every | |
1378 | # OSD id shows as many times as they are used in up and acting sets. | |
1379 | # If an OSD id is in both the up and acting set of a given PG, it will | |
1380 | # show twice. | |
1381 | # | |
1382 | # @param STDOUT a sorted list of OSD ids | |
1383 | # @return 0 on success, 1 on error | |
1384 | # | |
1385 | function get_osd_id_used_by_pgs() { | |
11fdf7f2 | 1386 | ceph --format json pg dump pgs 2>/dev/null | jq '.pg_stats | .[] | .up[], .acting[]' | sort |
c07f9fc5 FG |
1387 | } |
1388 | ||
1389 | function test_get_osd_id_used_by_pgs() { | |
1390 | local dir=$1 | |
1391 | ||
1392 | setup $dir || return 1 | |
1393 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1394 | run_mgr $dir x || return 1 | |
1395 | run_osd $dir 0 || return 1 | |
1396 | create_rbd_pool || return 1 | |
1397 | wait_for_clean || return 1 | |
1398 | local osd_ids=$(get_osd_id_used_by_pgs | uniq) | |
1399 | test "$osd_ids" = "0" || return 1 | |
1400 | teardown $dir || return 1 | |
1401 | } | |
1402 | ||
1403 | ####################################################################### | |
1404 | ||
1405 | ## | |
1406 | # Wait until the OSD **id** shows **count** times in the | |
1407 | # PGs (see get_osd_id_used_by_pgs for more information about | |
1408 | # how OSD ids are counted). | |
1409 | # | |
1410 | # @param id the OSD id | |
1411 | # @param count the number of time it must show in the PGs | |
1412 | # @return 0 on success, 1 on error | |
1413 | # | |
1414 | function wait_osd_id_used_by_pgs() { | |
1415 | local id=$1 | |
1416 | local count=$2 | |
1417 | ||
1418 | status=1 | |
1419 | for ((i=0; i < $TIMEOUT / 5; i++)); do | |
1420 | echo $i | |
1421 | if ! test $(get_osd_id_used_by_pgs | grep -c $id) = $count ; then | |
1422 | sleep 5 | |
1423 | else | |
1424 | status=0 | |
1425 | break | |
1426 | fi | |
1427 | done | |
1428 | return $status | |
1429 | } | |
1430 | ||
1431 | function test_wait_osd_id_used_by_pgs() { | |
1432 | local dir=$1 | |
1433 | ||
1434 | setup $dir || return 1 | |
1435 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1436 | run_mgr $dir x || return 1 | |
1437 | run_osd $dir 0 || return 1 | |
1438 | create_rbd_pool || return 1 | |
1439 | wait_for_clean || return 1 | |
1440 | wait_osd_id_used_by_pgs 0 8 || return 1 | |
1441 | ! TIMEOUT=1 wait_osd_id_used_by_pgs 123 5 || return 1 | |
1442 | teardown $dir || return 1 | |
1443 | } | |
1444 | ||
1445 | ####################################################################### | |
1446 | ||
7c673cae FG |
1447 | ## |
1448 | # Return the date and time of the last completed scrub for **pgid**, | |
1449 | # as reported by ceph pg dump pgs. Note that a repair also sets this | |
1450 | # date. | |
1451 | # | |
1452 | # @param pgid the id of the PG | |
1453 | # @param STDOUT the date and time of the last scrub | |
1454 | # @return 0 on success, 1 on error | |
1455 | # | |
1456 | function get_last_scrub_stamp() { | |
1457 | local pgid=$1 | |
1458 | local sname=${2:-last_scrub_stamp} | |
31f18b77 | 1459 | ceph --format json pg dump pgs 2>/dev/null | \ |
11fdf7f2 | 1460 | jq -r ".pg_stats | .[] | select(.pgid==\"$pgid\") | .$sname" |
7c673cae FG |
1461 | } |
1462 | ||
1463 | function test_get_last_scrub_stamp() { | |
1464 | local dir=$1 | |
1465 | ||
1466 | setup $dir || return 1 | |
1467 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1468 | run_mgr $dir x || return 1 | |
1469 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1470 | create_rbd_pool || return 1 |
7c673cae | 1471 | wait_for_clean || return 1 |
b5b8bbf5 | 1472 | stamp=$(get_last_scrub_stamp 1.0) |
7c673cae FG |
1473 | test -n "$stamp" || return 1 |
1474 | teardown $dir || return 1 | |
1475 | } | |
1476 | ||
1477 | ####################################################################### | |
1478 | ||
1479 | ## | |
1480 | # Predicate checking if the cluster is clean, i.e. all of its PGs are | |
1481 | # in a clean state (see get_num_active_clean for a definition). | |
1482 | # | |
1483 | # @return 0 if the cluster is clean, 1 otherwise | |
1484 | # | |
1485 | function is_clean() { | |
1486 | num_pgs=$(get_num_pgs) | |
1487 | test $num_pgs != 0 || return 1 | |
1488 | test $(get_num_active_clean) = $num_pgs || return 1 | |
1489 | } | |
1490 | ||
1491 | function test_is_clean() { | |
1492 | local dir=$1 | |
1493 | ||
1494 | setup $dir || return 1 | |
1495 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1496 | run_mgr $dir x || return 1 | |
1497 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1498 | create_rbd_pool || return 1 |
7c673cae FG |
1499 | wait_for_clean || return 1 |
1500 | is_clean || return 1 | |
1501 | teardown $dir || return 1 | |
1502 | } | |
1503 | ||
1504 | ####################################################################### | |
1505 | ||
f64942e4 | 1506 | calc() { $AWK "BEGIN{print $*}"; } |
94b18763 | 1507 | |
7c673cae FG |
1508 | ## |
1509 | # Return a list of numbers that are increasingly larger and whose | |
1510 | # total is **timeout** seconds. It can be used to have short sleep | |
1511 | # delay while waiting for an event on a fast machine. But if running | |
1512 | # very slowly the larger delays avoid stressing the machine even | |
1513 | # further or spamming the logs. | |
1514 | # | |
1515 | # @param timeout sum of all delays, in seconds | |
1516 | # @return a list of sleep delays | |
1517 | # | |
1518 | function get_timeout_delays() { | |
1519 | local trace=$(shopt -q -o xtrace && echo true || echo false) | |
1520 | $trace && shopt -u -o xtrace | |
1521 | local timeout=$1 | |
1522 | local first_step=${2:-1} | |
11fdf7f2 | 1523 | local max_timeout=${3:-$MAX_TIMEOUT} |
7c673cae FG |
1524 | |
1525 | local i | |
1526 | local total="0" | |
1527 | i=$first_step | |
94b18763 FG |
1528 | while test "$(calc $total + $i \<= $timeout)" = "1"; do |
1529 | echo -n "$(calc $i) " | |
1530 | total=$(calc $total + $i) | |
1531 | i=$(calc $i \* 2) | |
11fdf7f2 TL |
1532 | if [ $max_timeout -gt 0 ]; then |
1533 | # Did we reach max timeout ? | |
1534 | if [ ${i%.*} -eq ${max_timeout%.*} ] && [ ${i#*.} \> ${max_timeout#*.} ] || [ ${i%.*} -gt ${max_timeout%.*} ]; then | |
1535 | # Yes, so let's cap the max wait time to max | |
1536 | i=$max_timeout | |
1537 | fi | |
1538 | fi | |
7c673cae | 1539 | done |
94b18763 FG |
1540 | if test "$(calc $total \< $timeout)" = "1"; then |
1541 | echo -n "$(calc $timeout - $total) " | |
7c673cae FG |
1542 | fi |
1543 | $trace && shopt -s -o xtrace | |
1544 | } | |
1545 | ||
1546 | function test_get_timeout_delays() { | |
1547 | test "$(get_timeout_delays 1)" = "1 " || return 1 | |
94b18763 FG |
1548 | test "$(get_timeout_delays 5)" = "1 2 2 " || return 1 |
1549 | test "$(get_timeout_delays 6)" = "1 2 3 " || return 1 | |
7c673cae | 1550 | test "$(get_timeout_delays 7)" = "1 2 4 " || return 1 |
94b18763 FG |
1551 | test "$(get_timeout_delays 8)" = "1 2 4 1 " || return 1 |
1552 | test "$(get_timeout_delays 1 .1)" = "0.1 0.2 0.4 0.3 " || return 1 | |
1553 | test "$(get_timeout_delays 1.5 .1)" = "0.1 0.2 0.4 0.8 " || return 1 | |
1554 | test "$(get_timeout_delays 5 .1)" = "0.1 0.2 0.4 0.8 1.6 1.9 " || return 1 | |
1555 | test "$(get_timeout_delays 6 .1)" = "0.1 0.2 0.4 0.8 1.6 2.9 " || return 1 | |
1556 | test "$(get_timeout_delays 6.3 .1)" = "0.1 0.2 0.4 0.8 1.6 3.2 " || return 1 | |
1557 | test "$(get_timeout_delays 20 .1)" = "0.1 0.2 0.4 0.8 1.6 3.2 6.4 7.3 " || return 1 | |
11fdf7f2 TL |
1558 | test "$(get_timeout_delays 300 .1 0)" = "0.1 0.2 0.4 0.8 1.6 3.2 6.4 12.8 25.6 51.2 102.4 95.3 " || return 1 |
1559 | test "$(get_timeout_delays 300 .1 10)" = "0.1 0.2 0.4 0.8 1.6 3.2 6.4 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 7.3 " || return 1 | |
7c673cae FG |
1560 | } |
1561 | ||
1562 | ####################################################################### | |
1563 | ||
1564 | ## | |
1565 | # Wait until the cluster becomes clean or if it does not make progress | |
11fdf7f2 | 1566 | # for $WAIT_FOR_CLEAN_TIMEOUT seconds. |
7c673cae FG |
1567 | # Progress is measured either via the **get_is_making_recovery_progress** |
1568 | # predicate or if the number of clean PGs changes (as returned by get_num_active_clean) | |
1569 | # | |
1570 | # @return 0 if the cluster is clean, 1 otherwise | |
1571 | # | |
1572 | function wait_for_clean() { | |
a8e16298 | 1573 | local cmd=$1 |
7c673cae FG |
1574 | local num_active_clean=-1 |
1575 | local cur_active_clean | |
11fdf7f2 | 1576 | local -a delays=($(get_timeout_delays $WAIT_FOR_CLEAN_TIMEOUT .1)) |
7c673cae | 1577 | local -i loop=0 |
31f18b77 | 1578 | |
3a9019d9 | 1579 | flush_pg_stats || return 1 |
31f18b77 FG |
1580 | while test $(get_num_pgs) == 0 ; do |
1581 | sleep 1 | |
1582 | done | |
7c673cae FG |
1583 | |
1584 | while true ; do | |
1585 | # Comparing get_num_active_clean & get_num_pgs is used to determine | |
1586 | # if the cluster is clean. That's almost an inline of is_clean() to | |
1587 | # get more performance by avoiding multiple calls of get_num_active_clean. | |
1588 | cur_active_clean=$(get_num_active_clean) | |
1589 | test $cur_active_clean = $(get_num_pgs) && break | |
1590 | if test $cur_active_clean != $num_active_clean ; then | |
1591 | loop=0 | |
1592 | num_active_clean=$cur_active_clean | |
1593 | elif get_is_making_recovery_progress ; then | |
1594 | loop=0 | |
1595 | elif (( $loop >= ${#delays[*]} )) ; then | |
1596 | ceph report | |
1597 | return 1 | |
1598 | fi | |
a8e16298 TL |
1599 | # eval is a no-op if cmd is empty |
1600 | eval $cmd | |
7c673cae FG |
1601 | sleep ${delays[$loop]} |
1602 | loop+=1 | |
1603 | done | |
1604 | return 0 | |
1605 | } | |
1606 | ||
1607 | function test_wait_for_clean() { | |
1608 | local dir=$1 | |
1609 | ||
1610 | setup $dir || return 1 | |
11fdf7f2 TL |
1611 | run_mon $dir a --osd_pool_default_size=2 || return 1 |
1612 | run_osd $dir 0 || return 1 | |
7c673cae | 1613 | run_mgr $dir x || return 1 |
c07f9fc5 | 1614 | create_rbd_pool || return 1 |
11fdf7f2 TL |
1615 | ! WAIT_FOR_CLEAN_TIMEOUT=1 wait_for_clean || return 1 |
1616 | run_osd $dir 1 || return 1 | |
7c673cae FG |
1617 | wait_for_clean || return 1 |
1618 | teardown $dir || return 1 | |
1619 | } | |
1620 | ||
92f5a8d4 TL |
1621 | ## |
1622 | # Wait until the cluster becomes peered or if it does not make progress | |
1623 | # for $WAIT_FOR_CLEAN_TIMEOUT seconds. | |
1624 | # Progress is measured either via the **get_is_making_recovery_progress** | |
1625 | # predicate or if the number of peered PGs changes (as returned by get_num_active_or_peered) | |
1626 | # | |
1627 | # @return 0 if the cluster is clean, 1 otherwise | |
1628 | # | |
1629 | function wait_for_peered() { | |
1630 | local cmd=$1 | |
1631 | local num_peered=-1 | |
1632 | local cur_peered | |
1633 | local -a delays=($(get_timeout_delays $WAIT_FOR_CLEAN_TIMEOUT .1)) | |
1634 | local -i loop=0 | |
1635 | ||
1636 | flush_pg_stats || return 1 | |
1637 | while test $(get_num_pgs) == 0 ; do | |
1638 | sleep 1 | |
1639 | done | |
1640 | ||
1641 | while true ; do | |
1642 | # Comparing get_num_active_clean & get_num_pgs is used to determine | |
1643 | # if the cluster is clean. That's almost an inline of is_clean() to | |
1644 | # get more performance by avoiding multiple calls of get_num_active_clean. | |
1645 | cur_peered=$(get_num_active_or_peered) | |
1646 | test $cur_peered = $(get_num_pgs) && break | |
1647 | if test $cur_peered != $num_peered ; then | |
1648 | loop=0 | |
1649 | num_peered=$cur_peered | |
1650 | elif get_is_making_recovery_progress ; then | |
1651 | loop=0 | |
1652 | elif (( $loop >= ${#delays[*]} )) ; then | |
1653 | ceph report | |
1654 | return 1 | |
1655 | fi | |
1656 | # eval is a no-op if cmd is empty | |
1657 | eval $cmd | |
1658 | sleep ${delays[$loop]} | |
1659 | loop+=1 | |
1660 | done | |
1661 | return 0 | |
1662 | } | |
1663 | ||
1664 | function test_wait_for_peered() { | |
1665 | local dir=$1 | |
1666 | ||
1667 | setup $dir || return 1 | |
1668 | run_mon $dir a --osd_pool_default_size=2 || return 1 | |
1669 | run_osd $dir 0 || return 1 | |
1670 | run_mgr $dir x || return 1 | |
1671 | create_rbd_pool || return 1 | |
1672 | ! WAIT_FOR_CLEAN_TIMEOUT=1 wait_for_clean || return 1 | |
1673 | run_osd $dir 1 || return 1 | |
1674 | wait_for_peered || return 1 | |
1675 | teardown $dir || return 1 | |
1676 | } | |
1677 | ||
1678 | ||
7c673cae FG |
1679 | ####################################################################### |
1680 | ||
1681 | ## | |
1adf2230 AA |
1682 | # Wait until the cluster has health condition passed as arg |
1683 | # again for $TIMEOUT seconds. | |
7c673cae | 1684 | # |
1adf2230 AA |
1685 | # @param string to grep for in health detail |
1686 | # @return 0 if the cluster health matches request, 1 otherwise | |
7c673cae FG |
1687 | # |
1688 | function wait_for_health() { | |
1689 | local grepstr=$1 | |
1690 | local -a delays=($(get_timeout_delays $TIMEOUT .1)) | |
1691 | local -i loop=0 | |
1692 | ||
1693 | while ! ceph health detail | grep "$grepstr" ; do | |
1694 | if (( $loop >= ${#delays[*]} )) ; then | |
1695 | ceph health detail | |
1696 | return 1 | |
1697 | fi | |
1698 | sleep ${delays[$loop]} | |
1699 | loop+=1 | |
1700 | done | |
1701 | } | |
1702 | ||
1adf2230 AA |
1703 | ## |
1704 | # Wait until the cluster becomes HEALTH_OK again or if it does not make progress | |
1705 | # for $TIMEOUT seconds. | |
1706 | # | |
1707 | # @return 0 if the cluster is HEALTHY, 1 otherwise | |
1708 | # | |
7c673cae FG |
1709 | function wait_for_health_ok() { |
1710 | wait_for_health "HEALTH_OK" || return 1 | |
1711 | } | |
1712 | ||
1713 | function test_wait_for_health_ok() { | |
1714 | local dir=$1 | |
1715 | ||
1716 | setup $dir || return 1 | |
eafe8130 | 1717 | run_mon $dir a --osd_failsafe_full_ratio=.99 --mon_pg_warn_min_per_osd=0 || return 1 |
31f18b77 | 1718 | run_mgr $dir x --mon_pg_warn_min_per_osd=0 || return 1 |
eafe8130 | 1719 | # start osd_pool_default_size OSDs |
7c673cae | 1720 | run_osd $dir 0 || return 1 |
eafe8130 TL |
1721 | run_osd $dir 1 || return 1 |
1722 | run_osd $dir 2 || return 1 | |
224ce89b | 1723 | kill_daemons $dir TERM osd || return 1 |
11fdf7f2 | 1724 | ceph osd down 0 || return 1 |
eafe8130 | 1725 | # expect TOO_FEW_OSDS warning |
224ce89b | 1726 | ! TIMEOUT=1 wait_for_health_ok || return 1 |
eafe8130 | 1727 | # resurrect all OSDs |
224ce89b | 1728 | activate_osd $dir 0 || return 1 |
eafe8130 TL |
1729 | activate_osd $dir 1 || return 1 |
1730 | activate_osd $dir 2 || return 1 | |
7c673cae FG |
1731 | wait_for_health_ok || return 1 |
1732 | teardown $dir || return 1 | |
1733 | } | |
1734 | ||
1735 | ||
1736 | ####################################################################### | |
1737 | ||
1738 | ## | |
1739 | # Run repair on **pgid** and wait until it completes. The repair | |
1740 | # function will fail if repair does not complete within $TIMEOUT | |
1741 | # seconds. | |
1742 | # | |
1743 | # @param pgid the id of the PG | |
1744 | # @return 0 on success, 1 on error | |
1745 | # | |
1746 | function repair() { | |
1747 | local pgid=$1 | |
1748 | local last_scrub=$(get_last_scrub_stamp $pgid) | |
1749 | ceph pg repair $pgid | |
1750 | wait_for_scrub $pgid "$last_scrub" | |
1751 | } | |
1752 | ||
1753 | function test_repair() { | |
1754 | local dir=$1 | |
1755 | ||
1756 | setup $dir || return 1 | |
1757 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1758 | run_mgr $dir x || return 1 | |
1759 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1760 | create_rbd_pool || return 1 |
7c673cae | 1761 | wait_for_clean || return 1 |
b5b8bbf5 | 1762 | repair 1.0 || return 1 |
7c673cae | 1763 | kill_daemons $dir KILL osd || return 1 |
b5b8bbf5 | 1764 | ! TIMEOUT=1 repair 1.0 || return 1 |
7c673cae FG |
1765 | teardown $dir || return 1 |
1766 | } | |
1767 | ####################################################################### | |
1768 | ||
1769 | ## | |
1770 | # Run scrub on **pgid** and wait until it completes. The pg_scrub | |
1771 | # function will fail if repair does not complete within $TIMEOUT | |
1772 | # seconds. The pg_scrub is complete whenever the | |
1773 | # **get_last_scrub_stamp** function reports a timestamp different from | |
1774 | # the one stored before starting the scrub. | |
1775 | # | |
1776 | # @param pgid the id of the PG | |
1777 | # @return 0 on success, 1 on error | |
1778 | # | |
1779 | function pg_scrub() { | |
1780 | local pgid=$1 | |
1781 | local last_scrub=$(get_last_scrub_stamp $pgid) | |
1782 | ceph pg scrub $pgid | |
1783 | wait_for_scrub $pgid "$last_scrub" | |
1784 | } | |
1785 | ||
1786 | function pg_deep_scrub() { | |
1787 | local pgid=$1 | |
1788 | local last_scrub=$(get_last_scrub_stamp $pgid last_deep_scrub_stamp) | |
1789 | ceph pg deep-scrub $pgid | |
1790 | wait_for_scrub $pgid "$last_scrub" last_deep_scrub_stamp | |
1791 | } | |
1792 | ||
1793 | function test_pg_scrub() { | |
1794 | local dir=$1 | |
1795 | ||
1796 | setup $dir || return 1 | |
1797 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1798 | run_mgr $dir x || return 1 | |
1799 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1800 | create_rbd_pool || return 1 |
7c673cae | 1801 | wait_for_clean || return 1 |
b5b8bbf5 | 1802 | pg_scrub 1.0 || return 1 |
7c673cae | 1803 | kill_daemons $dir KILL osd || return 1 |
b5b8bbf5 | 1804 | ! TIMEOUT=1 pg_scrub 1.0 || return 1 |
7c673cae FG |
1805 | teardown $dir || return 1 |
1806 | } | |
1807 | ||
1808 | ####################################################################### | |
1809 | ||
1810 | ## | |
1811 | # Run the *command* and expect it to fail (i.e. return a non zero status). | |
1812 | # The output (stderr and stdout) is stored in a temporary file in *dir* | |
1813 | # and is expected to contain the string *expected*. | |
1814 | # | |
1815 | # Return 0 if the command failed and the string was found. Otherwise | |
1816 | # return 1 and cat the full output of the command on stderr for debug. | |
1817 | # | |
1818 | # @param dir temporary directory to store the output | |
1819 | # @param expected string to look for in the output | |
1820 | # @param command ... the command and its arguments | |
1821 | # @return 0 on success, 1 on error | |
1822 | # | |
1823 | ||
1824 | function expect_failure() { | |
1825 | local dir=$1 | |
1826 | shift | |
1827 | local expected="$1" | |
1828 | shift | |
1829 | local success | |
1830 | ||
1831 | if "$@" > $dir/out 2>&1 ; then | |
1832 | success=true | |
1833 | else | |
1834 | success=false | |
1835 | fi | |
1836 | ||
1837 | if $success || ! grep --quiet "$expected" $dir/out ; then | |
1838 | cat $dir/out >&2 | |
1839 | return 1 | |
1840 | else | |
1841 | return 0 | |
1842 | fi | |
1843 | } | |
1844 | ||
1845 | function test_expect_failure() { | |
1846 | local dir=$1 | |
1847 | ||
1848 | setup $dir || return 1 | |
1849 | expect_failure $dir FAIL bash -c 'echo FAIL ; exit 1' || return 1 | |
1850 | # the command did not fail | |
1851 | ! expect_failure $dir FAIL bash -c 'echo FAIL ; exit 0' > $dir/out || return 1 | |
1852 | grep --quiet FAIL $dir/out || return 1 | |
1853 | # the command failed but the output does not contain the expected string | |
1854 | ! expect_failure $dir FAIL bash -c 'echo UNEXPECTED ; exit 1' > $dir/out || return 1 | |
1855 | ! grep --quiet FAIL $dir/out || return 1 | |
1856 | teardown $dir || return 1 | |
1857 | } | |
1858 | ||
1859 | ####################################################################### | |
1860 | ||
1861 | ## | |
1862 | # Given the *last_scrub*, wait for scrub to happen on **pgid**. It | |
1863 | # will fail if scrub does not complete within $TIMEOUT seconds. The | |
1864 | # repair is complete whenever the **get_last_scrub_stamp** function | |
1865 | # reports a timestamp different from the one given in argument. | |
1866 | # | |
1867 | # @param pgid the id of the PG | |
1868 | # @param last_scrub timestamp of the last scrub for *pgid* | |
1869 | # @return 0 on success, 1 on error | |
1870 | # | |
1871 | function wait_for_scrub() { | |
1872 | local pgid=$1 | |
1873 | local last_scrub="$2" | |
1874 | local sname=${3:-last_scrub_stamp} | |
1875 | ||
1876 | for ((i=0; i < $TIMEOUT; i++)); do | |
b5b8bbf5 | 1877 | if test "$(get_last_scrub_stamp $pgid $sname)" '>' "$last_scrub" ; then |
7c673cae FG |
1878 | return 0 |
1879 | fi | |
1880 | sleep 1 | |
1881 | done | |
1882 | return 1 | |
1883 | } | |
1884 | ||
1885 | function test_wait_for_scrub() { | |
1886 | local dir=$1 | |
1887 | ||
1888 | setup $dir || return 1 | |
1889 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1890 | run_mgr $dir x || return 1 | |
1891 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1892 | create_rbd_pool || return 1 |
7c673cae | 1893 | wait_for_clean || return 1 |
b5b8bbf5 | 1894 | local pgid=1.0 |
7c673cae FG |
1895 | ceph pg repair $pgid |
1896 | local last_scrub=$(get_last_scrub_stamp $pgid) | |
1897 | wait_for_scrub $pgid "$last_scrub" || return 1 | |
1898 | kill_daemons $dir KILL osd || return 1 | |
1899 | last_scrub=$(get_last_scrub_stamp $pgid) | |
1900 | ! TIMEOUT=1 wait_for_scrub $pgid "$last_scrub" || return 1 | |
1901 | teardown $dir || return 1 | |
1902 | } | |
1903 | ||
1904 | ####################################################################### | |
1905 | ||
1906 | ## | |
1907 | # Return 0 if the erasure code *plugin* is available, 1 otherwise. | |
1908 | # | |
1909 | # @param plugin erasure code plugin | |
1910 | # @return 0 on success, 1 on error | |
1911 | # | |
1912 | ||
1913 | function erasure_code_plugin_exists() { | |
1914 | local plugin=$1 | |
1915 | local status | |
1916 | local grepstr | |
1917 | local s | |
1918 | case `uname` in | |
1919 | FreeBSD) grepstr="Cannot open.*$plugin" ;; | |
1920 | *) grepstr="$plugin.*No such file" ;; | |
1921 | esac | |
1922 | ||
1923 | s=$(ceph osd erasure-code-profile set TESTPROFILE plugin=$plugin 2>&1) | |
1924 | local status=$? | |
1925 | if [ $status -eq 0 ]; then | |
1926 | ceph osd erasure-code-profile rm TESTPROFILE | |
1927 | elif ! echo $s | grep --quiet "$grepstr" ; then | |
1928 | status=1 | |
1929 | # display why the string was rejected. | |
1930 | echo $s | |
1931 | fi | |
1932 | return $status | |
1933 | } | |
1934 | ||
1935 | function test_erasure_code_plugin_exists() { | |
1936 | local dir=$1 | |
1937 | ||
1938 | setup $dir || return 1 | |
1939 | run_mon $dir a || return 1 | |
1940 | run_mgr $dir x || return 1 | |
1941 | erasure_code_plugin_exists jerasure || return 1 | |
1942 | ! erasure_code_plugin_exists FAKE || return 1 | |
1943 | teardown $dir || return 1 | |
1944 | } | |
1945 | ||
1946 | ####################################################################### | |
1947 | ||
1948 | ## | |
1949 | # Display all log files from **dir** on stdout. | |
1950 | # | |
1951 | # @param dir directory in which all data is stored | |
1952 | # | |
1953 | ||
1954 | function display_logs() { | |
1955 | local dir=$1 | |
1956 | ||
1957 | find $dir -maxdepth 1 -name '*.log' | \ | |
1958 | while read file ; do | |
1959 | echo "======================= $file" | |
1960 | cat $file | |
1961 | done | |
1962 | } | |
1963 | ||
1964 | function test_display_logs() { | |
1965 | local dir=$1 | |
1966 | ||
1967 | setup $dir || return 1 | |
1968 | run_mon $dir a || return 1 | |
1969 | kill_daemons $dir || return 1 | |
1970 | display_logs $dir > $dir/log.out | |
1971 | grep --quiet mon.a.log $dir/log.out || return 1 | |
1972 | teardown $dir || return 1 | |
1973 | } | |
1974 | ||
1975 | ####################################################################### | |
1976 | ## | |
1977 | # Spawn a command in background and save the pid in the variable name | |
1978 | # passed in argument. To make the output reading easier, the output is | |
1979 | # prepend with the process id. | |
1980 | # | |
1981 | # Example: | |
1982 | # pids1="" | |
1983 | # run_in_background pids1 bash -c 'sleep 1; exit 1' | |
1984 | # | |
1985 | # @param pid_variable the variable name (not value) where the pids will be stored | |
1986 | # @param ... the command to execute | |
1987 | # @return only the pid_variable output should be considered and used with **wait_background** | |
1988 | # | |
1989 | function run_in_background() { | |
1990 | local pid_variable=$1 | |
94b18763 | 1991 | shift |
7c673cae | 1992 | # Execute the command and prepend the output with its pid |
f64942e4 | 1993 | # We enforce to return the exit status of the command and not the sed one. |
94b18763 | 1994 | ("$@" |& sed 's/^/'$$': /'; return "${PIPESTATUS[0]}") >&2 & |
7c673cae FG |
1995 | eval "$pid_variable+=\" $!\"" |
1996 | } | |
1997 | ||
94b18763 FG |
1998 | function save_stdout { |
1999 | local out="$1" | |
2000 | shift | |
2001 | "$@" > "$out" | |
2002 | } | |
2003 | ||
7c673cae FG |
2004 | function test_run_in_background() { |
2005 | local pids | |
2006 | run_in_background pids sleep 1 | |
2007 | run_in_background pids sleep 1 | |
2008 | test $(echo $pids | wc -w) = 2 || return 1 | |
2009 | wait $pids || return 1 | |
2010 | } | |
2011 | ||
2012 | ####################################################################### | |
2013 | ## | |
2014 | # Wait for pids running in background to complete. | |
2015 | # This function is usually used after a **run_in_background** call | |
2016 | # Example: | |
2017 | # pids1="" | |
2018 | # run_in_background pids1 bash -c 'sleep 1; exit 1' | |
2019 | # wait_background pids1 | |
2020 | # | |
2021 | # @param pids The variable name that contains the active PIDS. Set as empty at then end of the function. | |
2022 | # @return returns 1 if at least one process exits in error unless returns 0 | |
2023 | # | |
2024 | function wait_background() { | |
2025 | # We extract the PIDS from the variable name | |
2026 | pids=${!1} | |
2027 | ||
2028 | return_code=0 | |
2029 | for pid in $pids; do | |
2030 | if ! wait $pid; then | |
2031 | # If one process failed then return 1 | |
2032 | return_code=1 | |
2033 | fi | |
2034 | done | |
2035 | ||
2036 | # We empty the variable reporting that all process ended | |
2037 | eval "$1=''" | |
2038 | ||
2039 | return $return_code | |
2040 | } | |
2041 | ||
2042 | ||
2043 | function test_wait_background() { | |
2044 | local pids="" | |
2045 | run_in_background pids bash -c "sleep 1; exit 1" | |
2046 | run_in_background pids bash -c "sleep 2; exit 0" | |
2047 | wait_background pids | |
2048 | if [ $? -ne 1 ]; then return 1; fi | |
2049 | ||
2050 | run_in_background pids bash -c "sleep 1; exit 0" | |
2051 | run_in_background pids bash -c "sleep 2; exit 0" | |
2052 | wait_background pids | |
2053 | if [ $? -ne 0 ]; then return 1; fi | |
2054 | ||
2055 | if [ ! -z "$pids" ]; then return 1; fi | |
2056 | } | |
2057 | ||
31f18b77 FG |
2058 | function flush_pg_stats() |
2059 | { | |
2060 | local timeout=${1:-$TIMEOUT} | |
2061 | ||
2062 | ids=`ceph osd ls` | |
2063 | seqs='' | |
2064 | for osd in $ids; do | |
2065 | seq=`ceph tell osd.$osd flush_pg_stats` | |
2066 | seqs="$seqs $osd-$seq" | |
2067 | done | |
2068 | ||
2069 | for s in $seqs; do | |
2070 | osd=`echo $s | cut -d - -f 1` | |
2071 | seq=`echo $s | cut -d - -f 2` | |
2072 | echo "waiting osd.$osd seq $seq" | |
2073 | while test $(ceph osd last-stat-seq $osd) -lt $seq; do | |
2074 | sleep 1 | |
2075 | if [ $((timeout--)) -eq 0 ]; then | |
2076 | return 1 | |
2077 | fi | |
2078 | done | |
2079 | done | |
2080 | } | |
2081 | ||
2082 | function test_flush_pg_stats() | |
2083 | { | |
2084 | local dir=$1 | |
2085 | ||
2086 | setup $dir || return 1 | |
2087 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
2088 | run_mgr $dir x || return 1 | |
2089 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 2090 | create_rbd_pool || return 1 |
31f18b77 | 2091 | rados -p rbd put obj /etc/group |
3a9019d9 | 2092 | flush_pg_stats || return 1 |
31f18b77 | 2093 | local jq_filter='.pools | .[] | select(.name == "rbd") | .stats' |
11fdf7f2 TL |
2094 | stored=`ceph df detail --format=json | jq "$jq_filter.stored"` |
2095 | stored_raw=`ceph df detail --format=json | jq "$jq_filter.stored_raw"` | |
2096 | test $stored -gt 0 || return 1 | |
2097 | test $stored == $stored_raw || return 1 | |
b5b8bbf5 | 2098 | teardown $dir |
31f18b77 FG |
2099 | } |
2100 | ||
7c673cae FG |
2101 | ####################################################################### |
2102 | ||
2103 | ## | |
2104 | # Call the **run** function (which must be defined by the caller) with | |
2105 | # the **dir** argument followed by the caller argument list. | |
2106 | # | |
2107 | # If the **run** function returns on error, all logs found in **dir** | |
2108 | # are displayed for diagnostic purposes. | |
2109 | # | |
2110 | # **teardown** function is called when the **run** function returns | |
2111 | # (on success or on error), to cleanup leftovers. The CEPH_CONF is set | |
2112 | # to /dev/null and CEPH_ARGS is unset so that the tests are protected from | |
2113 | # external interferences. | |
2114 | # | |
2115 | # It is the responsibility of the **run** function to call the | |
2116 | # **setup** function to prepare the test environment (create a temporary | |
2117 | # directory etc.). | |
2118 | # | |
2119 | # The shell is required (via PS4) to display the function and line | |
2120 | # number whenever a statement is executed to help debugging. | |
2121 | # | |
2122 | # @param dir directory in which all data is stored | |
2123 | # @param ... arguments passed transparently to **run** | |
2124 | # @return 0 on success, 1 on error | |
2125 | # | |
2126 | function main() { | |
2127 | local dir=td/$1 | |
2128 | shift | |
2129 | ||
2130 | shopt -s -o xtrace | |
2131 | PS4='${BASH_SOURCE[0]}:$LINENO: ${FUNCNAME[0]}: ' | |
2132 | ||
11fdf7f2 TL |
2133 | export PATH=.:$PATH # make sure program from sources are preferred |
2134 | export PYTHONWARNINGS=ignore | |
7c673cae FG |
2135 | export CEPH_CONF=/dev/null |
2136 | unset CEPH_ARGS | |
2137 | ||
2138 | local code | |
2139 | if run $dir "$@" ; then | |
2140 | code=0 | |
2141 | else | |
7c673cae FG |
2142 | code=1 |
2143 | fi | |
b5b8bbf5 | 2144 | teardown $dir $code || return 1 |
7c673cae FG |
2145 | return $code |
2146 | } | |
2147 | ||
2148 | ####################################################################### | |
2149 | ||
2150 | function run_tests() { | |
2151 | shopt -s -o xtrace | |
2152 | PS4='${BASH_SOURCE[0]}:$LINENO: ${FUNCNAME[0]}: ' | |
2153 | ||
11fdf7f2 | 2154 | export .:$PATH # make sure program from sources are preferred |
7c673cae FG |
2155 | |
2156 | export CEPH_MON="127.0.0.1:7109" # git grep '\<7109\>' : there must be only one | |
2157 | export CEPH_ARGS | |
b5b8bbf5 | 2158 | CEPH_ARGS+=" --fsid=$(uuidgen) --auth-supported=none " |
7c673cae FG |
2159 | CEPH_ARGS+="--mon-host=$CEPH_MON " |
2160 | export CEPH_CONF=/dev/null | |
2161 | ||
2162 | local funcs=${@:-$(set | sed -n -e 's/^\(test_[0-9a-z_]*\) .*/\1/p')} | |
2163 | local dir=td/ceph-helpers | |
2164 | ||
2165 | for func in $funcs ; do | |
b5b8bbf5 FG |
2166 | if ! $func $dir; then |
2167 | teardown $dir 1 | |
2168 | return 1 | |
2169 | fi | |
7c673cae FG |
2170 | done |
2171 | } | |
2172 | ||
2173 | if test "$1" = TESTS ; then | |
2174 | shift | |
2175 | run_tests "$@" | |
b5b8bbf5 | 2176 | exit $? |
7c673cae FG |
2177 | fi |
2178 | ||
224ce89b WB |
2179 | # NOTE: |
2180 | # jq only support --exit-status|-e from version 1.4 forwards, which makes | |
2181 | # returning on error waaaay prettier and straightforward. | |
2182 | # However, the current automated upstream build is running with v1.3, | |
2183 | # which has no idea what -e is. Hence the convoluted error checking we | |
2184 | # need. Sad. | |
2185 | # The next time someone changes this code, please check if v1.4 is now | |
2186 | # a thing, and, if so, please change these to use -e. Thanks. | |
2187 | ||
2188 | # jq '.all.supported | select([.[] == "foo"] | any)' | |
2189 | function jq_success() { | |
2190 | input="$1" | |
2191 | filter="$2" | |
2192 | expects="\"$3\"" | |
2193 | ||
2194 | in_escaped=$(printf %s "$input" | sed "s/'/'\\\\''/g") | |
2195 | filter_escaped=$(printf %s "$filter" | sed "s/'/'\\\\''/g") | |
2196 | ||
2197 | ret=$(echo "$in_escaped" | jq "$filter_escaped") | |
2198 | if [[ "$ret" == "true" ]]; then | |
2199 | return 0 | |
2200 | elif [[ -n "$expects" ]]; then | |
2201 | if [[ "$ret" == "$expects" ]]; then | |
2202 | return 0 | |
2203 | fi | |
2204 | fi | |
2205 | return 1 | |
2206 | input=$1 | |
2207 | filter=$2 | |
2208 | expects="$3" | |
2209 | ||
2210 | ret="$(echo $input | jq \"$filter\")" | |
2211 | if [[ "$ret" == "true" ]]; then | |
2212 | return 0 | |
2213 | elif [[ -n "$expects" && "$ret" == "$expects" ]]; then | |
2214 | return 0 | |
2215 | fi | |
2216 | return 1 | |
2217 | } | |
2218 | ||
b5b8bbf5 FG |
2219 | function inject_eio() { |
2220 | local pooltype=$1 | |
2221 | shift | |
2222 | local which=$1 | |
2223 | shift | |
2224 | local poolname=$1 | |
2225 | shift | |
2226 | local objname=$1 | |
2227 | shift | |
2228 | local dir=$1 | |
2229 | shift | |
2230 | local shard_id=$1 | |
2231 | shift | |
2232 | ||
2233 | local -a initial_osds=($(get_osds $poolname $objname)) | |
2234 | local osd_id=${initial_osds[$shard_id]} | |
2235 | if [ "$pooltype" != "ec" ]; then | |
2236 | shard_id="" | |
2237 | fi | |
eafe8130 TL |
2238 | type=$(cat $dir/$osd_id/type) |
2239 | set_config osd $osd_id ${type}_debug_inject_read_err true || return 1 | |
b5b8bbf5 FG |
2240 | local loop=0 |
2241 | while ( CEPH_ARGS='' ceph --admin-daemon $(get_asok_path osd.$osd_id) \ | |
2242 | inject${which}err $poolname $objname $shard_id | grep -q Invalid ); do | |
2243 | loop=$(expr $loop + 1) | |
2244 | if [ $loop = "10" ]; then | |
2245 | return 1 | |
2246 | fi | |
2247 | sleep 1 | |
2248 | done | |
2249 | } | |
2250 | ||
1adf2230 AA |
2251 | function multidiff() { |
2252 | if ! diff $@ ; then | |
2253 | if [ "$DIFFCOLOPTS" = "" ]; then | |
2254 | return 1 | |
2255 | fi | |
2256 | diff $DIFFCOLOPTS $@ | |
2257 | fi | |
2258 | } | |
2259 | ||
eafe8130 TL |
2260 | function create_ec_pool() { |
2261 | local pool_name=$1 | |
2262 | shift | |
2263 | local allow_overwrites=$1 | |
2264 | shift | |
2265 | ||
2266 | ceph osd erasure-code-profile set myprofile crush-failure-domain=osd "$@" || return 1 | |
2267 | ||
2268 | create_pool "$poolname" 1 1 erasure myprofile || return 1 | |
2269 | ||
2270 | if [ "$allow_overwrites" = "true" ]; then | |
2271 | ceph osd pool set "$poolname" allow_ec_overwrites true || return 1 | |
2272 | fi | |
2273 | ||
2274 | wait_for_clean || return 1 | |
2275 | return 0 | |
2276 | } | |
2277 | ||
7c673cae | 2278 | # Local Variables: |
c07f9fc5 | 2279 | # compile-command: "cd ../../src ; make -j4 && ../qa/standalone/ceph-helpers.sh TESTS # test_get_config" |
7c673cae | 2280 | # End: |