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