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