]>
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 | |
22 | : ${CEPH_BUILD_VIRTUALENV:=/tmp} | |
23 | ||
24 | if type xmlstarlet > /dev/null 2>&1; then | |
25 | XMLSTARLET=xmlstarlet | |
26 | elif type xml > /dev/null 2>&1; then | |
27 | XMLSTARLET=xml | |
28 | else | |
29 | echo "Missing xmlstarlet binary!" | |
30 | exit 1 | |
31 | fi | |
31f18b77 | 32 | |
7c673cae FG |
33 | if [ `uname` = FreeBSD ]; then |
34 | SED=gsed | |
31f18b77 | 35 | DIFFCOLOPTS="" |
7c673cae FG |
36 | else |
37 | SED=sed | |
31f18b77 | 38 | termwidth=$(stty -a | head -1 | sed -e 's/.*columns \([0-9]*\).*/\1/') |
c07f9fc5 FG |
39 | if [ -n "$termwidth" -a "$termwidth" != "0" ]; then |
40 | termwidth="-W ${termwidth}" | |
31f18b77 FG |
41 | fi |
42 | DIFFCOLOPTS="-y $termwidth" | |
43 | fi | |
7c673cae | 44 | |
c07f9fc5 FG |
45 | EXTRA_OPTS="" |
46 | if [ -n "$CEPH_LIB" ]; then | |
47 | EXTRA_OPTS+=" --erasure-code-dir $CEPH_LIB" | |
48 | EXTRA_OPTS+=" --plugin-dir $CEPH_LIB" | |
49 | EXTRA_OPTS+=" --osd-class-dir $CEPH_LIB" | |
50 | fi | |
51 | ||
7c673cae FG |
52 | #! @file ceph-helpers.sh |
53 | # @brief Toolbox to manage Ceph cluster dedicated to testing | |
54 | # | |
55 | # Example use case: | |
56 | # | |
57 | # ~~~~~~~~~~~~~~~~{.sh} | |
58 | # source ceph-helpers.sh | |
59 | # | |
60 | # function mytest() { | |
61 | # # cleanup leftovers and reset mydir | |
62 | # setup mydir | |
63 | # # create a cluster with one monitor and three osds | |
64 | # run_mon mydir a | |
65 | # run_osd mydir 0 | |
66 | # run_osd mydir 2 | |
67 | # run_osd mydir 3 | |
68 | # # put and get an object | |
69 | # rados --pool rbd put GROUP /etc/group | |
70 | # rados --pool rbd get GROUP /tmp/GROUP | |
71 | # # stop the cluster and cleanup the directory | |
72 | # teardown mydir | |
73 | # } | |
74 | # ~~~~~~~~~~~~~~~~ | |
75 | # | |
76 | # The focus is on simplicity and efficiency, in the context of | |
77 | # functional tests. The output is intentionally very verbose | |
78 | # and functions return as soon as an error is found. The caller | |
79 | # is also expected to abort on the first error so that debugging | |
80 | # can be done by looking at the end of the output. | |
81 | # | |
82 | # Each function is documented, implemented and tested independently. | |
83 | # When modifying a helper, the test and the documentation are | |
84 | # expected to be updated and it is easier of they are collocated. A | |
85 | # test for a given function can be run with | |
86 | # | |
87 | # ~~~~~~~~~~~~~~~~{.sh} | |
88 | # ceph-helpers.sh TESTS test_get_osds | |
89 | # ~~~~~~~~~~~~~~~~ | |
90 | # | |
91 | # and all the tests (i.e. all functions matching test_*) are run | |
92 | # with: | |
93 | # | |
94 | # ~~~~~~~~~~~~~~~~{.sh} | |
95 | # ceph-helpers.sh TESTS | |
96 | # ~~~~~~~~~~~~~~~~ | |
97 | # | |
98 | # A test function takes a single argument : the directory dedicated | |
99 | # to the tests. It is expected to not create any file outside of this | |
100 | # directory and remove it entirely when it completes successfully. | |
101 | # | |
102 | ||
103 | ||
c07f9fc5 FG |
104 | function get_asok_dir() { |
105 | if [ -n "$CEPH_ASOK_DIR" ]; then | |
106 | echo "$CEPH_ASOK_DIR" | |
107 | else | |
108 | echo ${TMPDIR:-/tmp}/ceph-asok.$$ | |
109 | fi | |
110 | } | |
111 | ||
112 | function get_asok_path() { | |
113 | local name=$1 | |
114 | if [ -n "$name" ]; then | |
115 | echo $(get_asok_dir)/ceph-$name.asok | |
116 | else | |
117 | echo $(get_asok_dir)/\$cluster-\$name.asok | |
118 | fi | |
119 | } | |
7c673cae FG |
120 | ## |
121 | # Cleanup any leftovers found in **dir** via **teardown** | |
122 | # and reset **dir** as an empty environment. | |
123 | # | |
124 | # @param dir path name of the environment | |
125 | # @return 0 on success, 1 on error | |
126 | # | |
127 | function setup() { | |
128 | local dir=$1 | |
129 | teardown $dir || return 1 | |
130 | mkdir -p $dir | |
c07f9fc5 | 131 | mkdir -p $(get_asok_dir) |
7c673cae FG |
132 | } |
133 | ||
134 | function test_setup() { | |
135 | local dir=$dir | |
136 | setup $dir || return 1 | |
137 | test -d $dir || return 1 | |
138 | setup $dir || return 1 | |
139 | test -d $dir || return 1 | |
140 | teardown $dir | |
141 | } | |
142 | ||
143 | ####################################################################### | |
144 | ||
145 | ## | |
146 | # Kill all daemons for which a .pid file exists in **dir** and remove | |
147 | # **dir**. If the file system in which **dir** is btrfs, delete all | |
148 | # subvolumes that relate to it. | |
149 | # | |
150 | # @param dir path name of the environment | |
151 | # @return 0 on success, 1 on error | |
152 | # | |
153 | function teardown() { | |
154 | local dir=$1 | |
155 | kill_daemons $dir KILL | |
156 | if [ `uname` != FreeBSD ] \ | |
157 | && [ $(stat -f -c '%T' .) == "btrfs" ]; then | |
158 | __teardown_btrfs $dir | |
159 | fi | |
160 | rm -fr $dir | |
c07f9fc5 | 161 | rm -rf $(get_asok_dir) |
7c673cae FG |
162 | } |
163 | ||
164 | function __teardown_btrfs() { | |
165 | local btrfs_base_dir=$1 | |
166 | local btrfs_root=$(df -P . | tail -1 | awk '{print $NF}') | |
167 | local btrfs_dirs=$(cd $btrfs_base_dir; sudo btrfs subvolume list . -t | awk '/^[0-9]/ {print $4}' | grep "$btrfs_base_dir/$btrfs_dir") | |
168 | for subvolume in $btrfs_dirs; do | |
169 | sudo btrfs subvolume delete $btrfs_root/$subvolume | |
170 | done | |
171 | } | |
172 | ||
173 | function test_teardown() { | |
174 | local dir=$dir | |
175 | setup $dir || return 1 | |
176 | teardown $dir || return 1 | |
177 | ! test -d $dir || return 1 | |
178 | } | |
179 | ||
180 | ####################################################################### | |
181 | ||
182 | ## | |
183 | # Sends a signal to a single daemon. | |
184 | # This is a helper function for kill_daemons | |
185 | # | |
186 | # After the daemon is sent **signal**, its actual termination | |
187 | # will be verified by sending it signal 0. If the daemon is | |
188 | # still alive, kill_daemon will pause for a few seconds and | |
189 | # try again. This will repeat for a fixed number of times | |
190 | # before kill_daemon returns on failure. The list of | |
191 | # sleep intervals can be specified as **delays** and defaults | |
192 | # to: | |
193 | # | |
194 | # 0.1 0.2 1 1 1 2 3 5 5 5 10 10 20 60 60 60 120 | |
195 | # | |
196 | # This sequence is designed to run first a very short sleep time (0.1) | |
197 | # if the machine is fast enough and the daemon terminates in a fraction of a | |
198 | # second. The increasing sleep numbers should give plenty of time for | |
199 | # the daemon to die even on the slowest running machine. If a daemon | |
200 | # takes more than a few minutes to stop (the sum of all sleep times), | |
201 | # there probably is no point in waiting more and a number of things | |
202 | # are likely to go wrong anyway: better give up and return on error. | |
203 | # | |
204 | # @param pid the process id to send a signal | |
205 | # @param send_signal the signal to send | |
206 | # @param delays sequence of sleep times before failure | |
207 | # | |
208 | function kill_daemon() { | |
7c673cae FG |
209 | local pid=$(cat $1) |
210 | local send_signal=$2 | |
211 | local delays=${3:-0.1 0.2 1 1 1 2 3 5 5 5 10 10 20 60 60 60 120} | |
212 | local exit_code=1 | |
213 | for try in $delays ; do | |
214 | if kill -$send_signal $pid 2> /dev/null ; then | |
215 | exit_code=1 | |
216 | else | |
217 | exit_code=0 | |
218 | break | |
219 | fi | |
220 | send_signal=0 | |
221 | sleep $try | |
222 | done; | |
223 | return $exit_code | |
224 | } | |
225 | ||
226 | function test_kill_daemon() { | |
227 | local dir=$1 | |
228 | setup $dir || return 1 | |
229 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
230 | run_mgr $dir x || return 1 | |
231 | run_osd $dir 0 || return 1 | |
232 | ||
233 | name_prefix=osd | |
234 | for pidfile in $(find $dir 2>/dev/null | grep $name_prefix'[^/]*\.pid') ; do | |
235 | # | |
236 | # sending signal 0 won't kill the daemon | |
237 | # waiting just for one second instead of the default schedule | |
238 | # allows us to quickly verify what happens when kill fails | |
239 | # to stop the daemon (i.e. it must return false) | |
240 | # | |
241 | ! kill_daemon $pidfile 0 1 || return 1 | |
242 | # | |
243 | # killing just the osd and verify the mon still is responsive | |
244 | # | |
245 | kill_daemon $pidfile TERM || return 1 | |
246 | done | |
247 | ||
248 | ceph osd dump | grep "osd.0 down" || return 1 | |
249 | ||
250 | name_prefix=mgr | |
251 | for pidfile in $(find $dir 2>/dev/null | grep $name_prefix'[^/]*\.pid') ; do | |
252 | # | |
253 | # kill the mgr | |
254 | # | |
255 | kill_daemon $pidfile TERM || return 1 | |
256 | done | |
257 | ||
258 | name_prefix=mon | |
259 | for pidfile in $(find $dir 2>/dev/null | grep $name_prefix'[^/]*\.pid') ; do | |
260 | # | |
261 | # kill the mon and verify it cannot be reached | |
262 | # | |
263 | kill_daemon $pidfile TERM || return 1 | |
224ce89b | 264 | ! timeout 5 ceph status || return 1 |
7c673cae FG |
265 | done |
266 | ||
267 | teardown $dir || return 1 | |
268 | } | |
269 | ||
270 | ## | |
271 | # Kill all daemons for which a .pid file exists in **dir**. Each | |
272 | # daemon is sent a **signal** and kill_daemons waits for it to exit | |
273 | # during a few minutes. By default all daemons are killed. If a | |
274 | # **name_prefix** is provided, only the daemons for which a pid | |
275 | # file is found matching the prefix are killed. See run_osd and | |
276 | # run_mon for more information about the name conventions for | |
277 | # the pid files. | |
278 | # | |
279 | # Send TERM to all daemons : kill_daemons $dir | |
280 | # Send KILL to all daemons : kill_daemons $dir KILL | |
281 | # Send KILL to all osds : kill_daemons $dir KILL osd | |
282 | # Send KILL to osd 1 : kill_daemons $dir KILL osd.1 | |
283 | # | |
284 | # If a daemon is sent the TERM signal and does not terminate | |
285 | # within a few minutes, it will still be running even after | |
c07f9fc5 | 286 | # kill_daemons returns. |
7c673cae FG |
287 | # |
288 | # If all daemons are kill successfully the function returns 0 | |
c07f9fc5 | 289 | # if at least one daemon remains, this is treated as an |
7c673cae FG |
290 | # error and the function return 1. |
291 | # | |
292 | # @param dir path name of the environment | |
293 | # @param signal name of the first signal (defaults to TERM) | |
294 | # @param name_prefix only kill match daemons (defaults to all) | |
295 | # @param delays sequence of sleep times before failure | |
296 | # @return 0 on success, 1 on error | |
297 | # | |
298 | function kill_daemons() { | |
299 | local trace=$(shopt -q -o xtrace && echo true || echo false) | |
300 | $trace && shopt -u -o xtrace | |
301 | local dir=$1 | |
302 | local signal=${2:-TERM} | |
303 | local name_prefix=$3 # optional, osd, mon, osd.1 | |
304 | local delays=$4 #optional timing | |
305 | local status=0 | |
306 | local pids="" | |
307 | ||
308 | for pidfile in $(find $dir 2>/dev/null | grep $name_prefix'[^/]*\.pid') ; do | |
309 | run_in_background pids kill_daemon $pidfile $signal $delays | |
310 | done | |
311 | ||
312 | wait_background pids | |
313 | status=$? | |
314 | ||
315 | $trace && shopt -s -o xtrace | |
316 | return $status | |
317 | } | |
318 | ||
319 | function test_kill_daemons() { | |
320 | local dir=$1 | |
321 | setup $dir || return 1 | |
322 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
323 | run_mgr $dir x || return 1 | |
324 | run_osd $dir 0 || return 1 | |
325 | # | |
326 | # sending signal 0 won't kill the daemon | |
327 | # waiting just for one second instead of the default schedule | |
c07f9fc5 | 328 | # allows us to quickly verify what happens when kill fails |
7c673cae FG |
329 | # to stop the daemon (i.e. it must return false) |
330 | # | |
331 | ! kill_daemons $dir 0 osd 1 || return 1 | |
332 | # | |
333 | # killing just the osd and verify the mon still is responsive | |
334 | # | |
335 | kill_daemons $dir TERM osd || return 1 | |
336 | ceph osd dump | grep "osd.0 down" || return 1 | |
337 | # | |
338 | # kill the mgr | |
339 | # | |
340 | kill_daemons $dir TERM mgr || return 1 | |
341 | # | |
342 | # kill the mon and verify it cannot be reached | |
343 | # | |
344 | kill_daemons $dir TERM || return 1 | |
224ce89b | 345 | ! timeout 5 ceph status || return 1 |
7c673cae FG |
346 | teardown $dir || return 1 |
347 | } | |
348 | ||
349 | ####################################################################### | |
350 | ||
351 | ## | |
352 | # Run a monitor by the name mon.**id** with data in **dir**/**id**. | |
353 | # The logs can be found in **dir**/mon.**id**.log and the pid file | |
354 | # is **dir**/mon.**id**.pid and the admin socket is | |
355 | # **dir**/**id**/ceph-mon.**id**.asok. | |
356 | # | |
357 | # The remaining arguments are passed verbatim to ceph-mon --mkfs | |
358 | # and the ceph-mon daemon. | |
359 | # | |
360 | # Two mandatory arguments must be provided: --fsid and --mon-host | |
361 | # Instead of adding them to every call to run_mon, they can be | |
362 | # set in the CEPH_ARGS environment variable to be read implicitly | |
363 | # by every ceph command. | |
364 | # | |
365 | # The CEPH_CONF variable is expected to be set to /dev/null to | |
366 | # only rely on arguments for configuration. | |
367 | # | |
368 | # Examples: | |
369 | # | |
370 | # CEPH_ARGS="--fsid=$(uuidgen) " | |
371 | # CEPH_ARGS+="--mon-host=127.0.0.1:7018 " | |
372 | # run_mon $dir a # spawn a mon and bind port 7018 | |
373 | # run_mon $dir a --debug-filestore=20 # spawn with filestore debugging | |
374 | # | |
375 | # If mon_initial_members is not set, the default rbd pool is deleted | |
376 | # and replaced with a replicated pool with less placement groups to | |
377 | # speed up initialization. If mon_initial_members is set, no attempt | |
378 | # is made to recreate the rbd pool because it would hang forever, | |
379 | # waiting for other mons to join. | |
380 | # | |
381 | # A **dir**/ceph.conf file is created but not meant to be used by any | |
382 | # function. It is convenient for debugging a failure with: | |
383 | # | |
384 | # ceph --conf **dir**/ceph.conf -s | |
385 | # | |
386 | # @param dir path name of the environment | |
387 | # @param id mon identifier | |
388 | # @param ... can be any option valid for ceph-mon | |
389 | # @return 0 on success, 1 on error | |
390 | # | |
c07f9fc5 | 391 | function run_mon() { |
7c673cae FG |
392 | local dir=$1 |
393 | shift | |
394 | local id=$1 | |
395 | shift | |
396 | local data=$dir/$id | |
397 | ||
398 | ceph-mon \ | |
399 | --id $id \ | |
400 | --mkfs \ | |
401 | --mon-data=$data \ | |
402 | --run-dir=$dir \ | |
403 | "$@" || return 1 | |
404 | ||
405 | ceph-mon \ | |
406 | --id $id \ | |
407 | --mon-osd-full-ratio=.99 \ | |
408 | --mon-data-avail-crit=1 \ | |
409 | --paxos-propose-interval=0.1 \ | |
410 | --osd-crush-chooseleaf-type=0 \ | |
c07f9fc5 | 411 | $EXTRA_OPTS \ |
7c673cae FG |
412 | --debug-mon 20 \ |
413 | --debug-ms 20 \ | |
414 | --debug-paxos 20 \ | |
415 | --chdir= \ | |
416 | --mon-data=$data \ | |
417 | --log-file=$dir/\$name.log \ | |
c07f9fc5 | 418 | --admin-socket=$(get_asok_path) \ |
7c673cae FG |
419 | --mon-cluster-log-file=$dir/log \ |
420 | --run-dir=$dir \ | |
421 | --pid-file=$dir/\$name.pid \ | |
422 | --mon-allow-pool-delete \ | |
c07f9fc5 | 423 | --mon-osd-backfillfull-ratio .99 \ |
7c673cae FG |
424 | "$@" || return 1 |
425 | ||
426 | cat > $dir/ceph.conf <<EOF | |
427 | [global] | |
428 | fsid = $(get_config mon $id fsid) | |
429 | mon host = $(get_config mon $id mon_host) | |
430 | EOF | |
224ce89b WB |
431 | } |
432 | ||
7c673cae FG |
433 | function test_run_mon() { |
434 | local dir=$1 | |
435 | ||
436 | setup $dir || return 1 | |
437 | ||
438 | run_mon $dir a --mon-initial-members=a || return 1 | |
c07f9fc5 | 439 | create_rbd_pool || return 1 |
7c673cae | 440 | # rbd has not been deleted / created, hence it has pool id 0 |
224ce89b | 441 | ceph osd dump | grep "pool 1 'rbd'" || return 1 |
7c673cae FG |
442 | kill_daemons $dir || return 1 |
443 | ||
444 | run_mon $dir a || return 1 | |
c07f9fc5 | 445 | create_rbd_pool || return 1 |
7c673cae | 446 | # rbd has been deleted / created, hence it does not have pool id 0 |
224ce89b | 447 | ! ceph osd dump | grep "pool 1 'rbd'" || return 1 |
c07f9fc5 | 448 | local size=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path mon.a) \ |
7c673cae FG |
449 | config get osd_pool_default_size) |
450 | test "$size" = '{"osd_pool_default_size":"3"}' || return 1 | |
451 | ||
452 | ! CEPH_ARGS='' ceph status || return 1 | |
453 | CEPH_ARGS='' ceph --conf $dir/ceph.conf status || return 1 | |
454 | ||
455 | kill_daemons $dir || return 1 | |
456 | ||
457 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
c07f9fc5 | 458 | local size=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path mon.a) \ |
7c673cae FG |
459 | config get osd_pool_default_size) |
460 | test "$size" = '{"osd_pool_default_size":"1"}' || return 1 | |
461 | kill_daemons $dir || return 1 | |
462 | ||
463 | CEPH_ARGS="$CEPH_ARGS --osd_pool_default_size=2" \ | |
464 | run_mon $dir a || return 1 | |
c07f9fc5 | 465 | local size=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path mon.a) \ |
7c673cae FG |
466 | config get osd_pool_default_size) |
467 | test "$size" = '{"osd_pool_default_size":"2"}' || return 1 | |
468 | kill_daemons $dir || return 1 | |
469 | ||
470 | teardown $dir || return 1 | |
471 | } | |
472 | ||
c07f9fc5 FG |
473 | function create_rbd_pool() { |
474 | ceph osd pool delete rbd rbd --yes-i-really-really-mean-it || return 1 | |
475 | ceph osd pool create rbd $PG_NUM || return 1 | |
476 | rbd pool init rbd | |
477 | } | |
478 | ||
7c673cae FG |
479 | ####################################################################### |
480 | ||
481 | function run_mgr() { | |
482 | local dir=$1 | |
483 | shift | |
484 | local id=$1 | |
485 | shift | |
486 | local data=$dir/$id | |
487 | ||
488 | ceph-mgr \ | |
489 | --id $id \ | |
c07f9fc5 | 490 | $EXTRA_OPTS \ |
7c673cae FG |
491 | --debug-mgr 20 \ |
492 | --debug-objecter 20 \ | |
493 | --debug-ms 20 \ | |
494 | --debug-paxos 20 \ | |
495 | --chdir= \ | |
496 | --mgr-data=$data \ | |
497 | --log-file=$dir/\$name.log \ | |
c07f9fc5 | 498 | --admin-socket=$(get_asok_path) \ |
7c673cae FG |
499 | --run-dir=$dir \ |
500 | --pid-file=$dir/\$name.pid \ | |
501 | "$@" || return 1 | |
502 | } | |
503 | ||
504 | ####################################################################### | |
505 | ||
506 | ## | |
507 | # Create (prepare) and run (activate) an osd by the name osd.**id** | |
508 | # with data in **dir**/**id**. The logs can be found in | |
509 | # **dir**/osd.**id**.log, the pid file is **dir**/osd.**id**.pid and | |
510 | # the admin socket is **dir**/**id**/ceph-osd.**id**.asok. | |
511 | # | |
512 | # The remaining arguments are passed verbatim to ceph-osd. | |
513 | # | |
514 | # Two mandatory arguments must be provided: --fsid and --mon-host | |
515 | # Instead of adding them to every call to run_osd, they can be | |
516 | # set in the CEPH_ARGS environment variable to be read implicitly | |
517 | # by every ceph command. | |
518 | # | |
519 | # The CEPH_CONF variable is expected to be set to /dev/null to | |
520 | # only rely on arguments for configuration. | |
521 | # | |
522 | # The run_osd function creates the OSD data directory with ceph-disk | |
523 | # prepare on the **dir**/**id** directory and relies on the | |
524 | # activate_osd function to run the daemon. | |
525 | # | |
526 | # Examples: | |
527 | # | |
528 | # CEPH_ARGS="--fsid=$(uuidgen) " | |
529 | # CEPH_ARGS+="--mon-host=127.0.0.1:7018 " | |
530 | # run_osd $dir 0 # prepare and activate an osd using the monitor listening on 7018 | |
531 | # | |
532 | # @param dir path name of the environment | |
533 | # @param id osd identifier | |
534 | # @param ... can be any option valid for ceph-osd | |
535 | # @return 0 on success, 1 on error | |
536 | # | |
537 | function run_osd() { | |
538 | local dir=$1 | |
539 | shift | |
540 | local id=$1 | |
541 | shift | |
542 | local osd_data=$dir/$id | |
543 | ||
544 | local ceph_disk_args | |
545 | ceph_disk_args+=" --statedir=$dir" | |
546 | ceph_disk_args+=" --sysconfdir=$dir" | |
547 | ceph_disk_args+=" --prepend-to-path=" | |
548 | ||
549 | mkdir -p $osd_data | |
550 | ceph-disk $ceph_disk_args \ | |
31f18b77 | 551 | prepare --filestore $osd_data || return 1 |
7c673cae FG |
552 | |
553 | activate_osd $dir $id "$@" | |
554 | } | |
555 | ||
556 | function run_osd_bluestore() { | |
557 | local dir=$1 | |
558 | shift | |
559 | local id=$1 | |
560 | shift | |
561 | local osd_data=$dir/$id | |
562 | ||
563 | local ceph_disk_args | |
564 | ceph_disk_args+=" --statedir=$dir" | |
565 | ceph_disk_args+=" --sysconfdir=$dir" | |
566 | ceph_disk_args+=" --prepend-to-path=" | |
567 | ||
568 | mkdir -p $osd_data | |
569 | ceph-disk $ceph_disk_args \ | |
570 | prepare --bluestore $osd_data || return 1 | |
571 | ||
31f18b77 | 572 | activate_osd $dir $id "$@" |
7c673cae FG |
573 | } |
574 | ||
575 | function test_run_osd() { | |
576 | local dir=$1 | |
577 | ||
578 | setup $dir || return 1 | |
579 | ||
580 | run_mon $dir a || return 1 | |
581 | run_mgr $dir x || return 1 | |
582 | ||
583 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 584 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path osd.0) \ |
7c673cae FG |
585 | config get osd_max_backfills) |
586 | echo "$backfills" | grep --quiet 'osd_max_backfills' || return 1 | |
587 | ||
588 | run_osd $dir 1 --osd-max-backfills 20 || return 1 | |
c07f9fc5 | 589 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path osd.1) \ |
7c673cae FG |
590 | config get osd_max_backfills) |
591 | test "$backfills" = '{"osd_max_backfills":"20"}' || return 1 | |
592 | ||
593 | CEPH_ARGS="$CEPH_ARGS --osd-max-backfills 30" run_osd $dir 2 || return 1 | |
c07f9fc5 | 594 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path osd.2) \ |
7c673cae FG |
595 | config get osd_max_backfills) |
596 | test "$backfills" = '{"osd_max_backfills":"30"}' || return 1 | |
597 | ||
598 | teardown $dir || return 1 | |
599 | } | |
600 | ||
601 | ####################################################################### | |
602 | ||
603 | ## | |
604 | # Shutdown and remove all traces of the osd by the name osd.**id**. | |
605 | # | |
606 | # The OSD is shutdown with the TERM signal. It is then removed from | |
607 | # the auth list, crush map, osd map etc and the files associated with | |
608 | # it are also removed. | |
609 | # | |
610 | # @param dir path name of the environment | |
611 | # @param id osd identifier | |
612 | # @return 0 on success, 1 on error | |
613 | # | |
614 | function destroy_osd() { | |
615 | local dir=$1 | |
616 | local id=$2 | |
617 | ||
7c673cae | 618 | ceph osd out osd.$id || return 1 |
c07f9fc5 FG |
619 | kill_daemons $dir TERM osd.$id || return 1 |
620 | ceph osd purge osd.$id --yes-i-really-mean-it || return 1 | |
7c673cae FG |
621 | teardown $dir/$id || return 1 |
622 | rm -fr $dir/$id | |
623 | } | |
624 | ||
625 | function test_destroy_osd() { | |
626 | local dir=$1 | |
627 | ||
628 | setup $dir || return 1 | |
629 | run_mon $dir a || return 1 | |
630 | run_mgr $dir x || return 1 | |
631 | run_osd $dir 0 || return 1 | |
632 | destroy_osd $dir 0 || return 1 | |
633 | ! ceph osd dump | grep "osd.$id " || return 1 | |
634 | teardown $dir || return 1 | |
635 | } | |
636 | ||
637 | ####################################################################### | |
638 | ||
639 | ## | |
640 | # Run (activate) an osd by the name osd.**id** with data in | |
641 | # **dir**/**id**. The logs can be found in **dir**/osd.**id**.log, | |
642 | # the pid file is **dir**/osd.**id**.pid and the admin socket is | |
643 | # **dir**/**id**/ceph-osd.**id**.asok. | |
644 | # | |
645 | # The remaining arguments are passed verbatim to ceph-osd. | |
646 | # | |
647 | # Two mandatory arguments must be provided: --fsid and --mon-host | |
648 | # Instead of adding them to every call to activate_osd, they can be | |
649 | # set in the CEPH_ARGS environment variable to be read implicitly | |
650 | # by every ceph command. | |
651 | # | |
652 | # The CEPH_CONF variable is expected to be set to /dev/null to | |
653 | # only rely on arguments for configuration. | |
654 | # | |
655 | # The activate_osd function expects a valid OSD data directory | |
656 | # in **dir**/**id**, either just created via run_osd or re-using | |
657 | # one left by a previous run of ceph-osd. The ceph-osd daemon is | |
658 | # run indirectly via ceph-disk activate. | |
659 | # | |
660 | # The activate_osd function blocks until the monitor reports the osd | |
661 | # up. If it fails to do so within $TIMEOUT seconds, activate_osd | |
662 | # fails. | |
663 | # | |
664 | # Examples: | |
665 | # | |
666 | # CEPH_ARGS="--fsid=$(uuidgen) " | |
667 | # CEPH_ARGS+="--mon-host=127.0.0.1:7018 " | |
668 | # activate_osd $dir 0 # activate an osd using the monitor listening on 7018 | |
669 | # | |
670 | # @param dir path name of the environment | |
671 | # @param id osd identifier | |
672 | # @param ... can be any option valid for ceph-osd | |
673 | # @return 0 on success, 1 on error | |
674 | # | |
675 | function activate_osd() { | |
676 | local dir=$1 | |
677 | shift | |
678 | local id=$1 | |
679 | shift | |
680 | local osd_data=$dir/$id | |
681 | ||
682 | local ceph_disk_args | |
683 | ceph_disk_args+=" --statedir=$dir" | |
684 | ceph_disk_args+=" --sysconfdir=$dir" | |
685 | ceph_disk_args+=" --prepend-to-path=" | |
686 | ||
687 | local ceph_args="$CEPH_ARGS" | |
7c673cae FG |
688 | ceph_args+=" --osd-failsafe-full-ratio=.99" |
689 | ceph_args+=" --osd-journal-size=100" | |
690 | ceph_args+=" --osd-scrub-load-threshold=2000" | |
691 | ceph_args+=" --osd-data=$osd_data" | |
692 | ceph_args+=" --chdir=" | |
c07f9fc5 | 693 | ceph_args+=$EXTRA_OPTS |
7c673cae | 694 | ceph_args+=" --run-dir=$dir" |
c07f9fc5 | 695 | ceph_args+=" --admin-socket=$(get_asok_path)" |
7c673cae FG |
696 | ceph_args+=" --debug-osd=20" |
697 | ceph_args+=" --log-file=$dir/\$name.log" | |
698 | ceph_args+=" --pid-file=$dir/\$name.pid" | |
699 | ceph_args+=" --osd-max-object-name-len 460" | |
700 | ceph_args+=" --osd-max-object-namespace-len 64" | |
224ce89b | 701 | ceph_args+=" --enable-experimental-unrecoverable-data-corrupting-features *" |
7c673cae FG |
702 | ceph_args+=" " |
703 | ceph_args+="$@" | |
704 | mkdir -p $osd_data | |
705 | CEPH_ARGS="$ceph_args " ceph-disk $ceph_disk_args \ | |
706 | activate \ | |
707 | --mark-init=none \ | |
708 | $osd_data || return 1 | |
709 | ||
710 | [ "$id" = "$(cat $osd_data/whoami)" ] || return 1 | |
711 | ||
712 | wait_for_osd up $id || return 1 | |
713 | } | |
714 | ||
715 | function test_activate_osd() { | |
716 | local dir=$1 | |
717 | ||
718 | setup $dir || return 1 | |
719 | ||
720 | run_mon $dir a || return 1 | |
721 | run_mgr $dir x || return 1 | |
722 | ||
723 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 724 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path osd.0) \ |
7c673cae FG |
725 | config get osd_max_backfills) |
726 | echo "$backfills" | grep --quiet 'osd_max_backfills' || return 1 | |
727 | ||
728 | kill_daemons $dir TERM osd || return 1 | |
729 | ||
730 | activate_osd $dir 0 --osd-max-backfills 20 || return 1 | |
c07f9fc5 | 731 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $(get_asok_path osd.0) \ |
7c673cae FG |
732 | config get osd_max_backfills) |
733 | test "$backfills" = '{"osd_max_backfills":"20"}' || return 1 | |
734 | ||
735 | teardown $dir || return 1 | |
736 | } | |
737 | ||
738 | ####################################################################### | |
739 | ||
740 | ## | |
741 | # Wait until the OSD **id** is either up or down, as specified by | |
742 | # **state**. It fails after $TIMEOUT seconds. | |
743 | # | |
744 | # @param state either up or down | |
745 | # @param id osd identifier | |
746 | # @return 0 on success, 1 on error | |
747 | # | |
748 | function wait_for_osd() { | |
749 | local state=$1 | |
750 | local id=$2 | |
751 | ||
752 | status=1 | |
753 | for ((i=0; i < $TIMEOUT; i++)); do | |
754 | echo $i | |
755 | if ! ceph osd dump | grep "osd.$id $state"; then | |
756 | sleep 1 | |
757 | else | |
758 | status=0 | |
759 | break | |
760 | fi | |
761 | done | |
762 | return $status | |
763 | } | |
764 | ||
765 | function test_wait_for_osd() { | |
766 | local dir=$1 | |
767 | setup $dir || return 1 | |
768 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
769 | run_mgr $dir x || return 1 | |
770 | run_osd $dir 0 || return 1 | |
771 | wait_for_osd up 0 || return 1 | |
772 | kill_daemons $dir TERM osd || return 1 | |
773 | wait_for_osd down 0 || return 1 | |
774 | ( TIMEOUT=1 ; ! wait_for_osd up 0 ) || return 1 | |
775 | teardown $dir || return 1 | |
776 | } | |
777 | ||
778 | ####################################################################### | |
779 | ||
780 | ## | |
781 | # Display the list of OSD ids supporting the **objectname** stored in | |
782 | # **poolname**, as reported by ceph osd map. | |
783 | # | |
784 | # @param poolname an existing pool | |
785 | # @param objectname an objectname (may or may not exist) | |
786 | # @param STDOUT white space separated list of OSD ids | |
787 | # @return 0 on success, 1 on error | |
788 | # | |
789 | function get_osds() { | |
790 | local poolname=$1 | |
791 | local objectname=$2 | |
792 | ||
31f18b77 FG |
793 | local osds=$(ceph --format json osd map $poolname $objectname 2>/dev/null | \ |
794 | jq '.acting | .[]') | |
7c673cae FG |
795 | # get rid of the trailing space |
796 | echo $osds | |
797 | } | |
798 | ||
799 | function test_get_osds() { | |
800 | local dir=$1 | |
801 | ||
802 | setup $dir || return 1 | |
803 | run_mon $dir a --osd_pool_default_size=2 || return 1 | |
804 | run_mgr $dir x || return 1 | |
805 | run_osd $dir 0 || return 1 | |
806 | run_osd $dir 1 || return 1 | |
c07f9fc5 | 807 | create_rbd_pool || return 1 |
7c673cae | 808 | wait_for_clean || return 1 |
c07f9fc5 | 809 | create_rbd_pool || return 1 |
7c673cae FG |
810 | get_osds rbd GROUP | grep --quiet '^[0-1] [0-1]$' || return 1 |
811 | teardown $dir || return 1 | |
812 | } | |
813 | ||
814 | ####################################################################### | |
815 | ||
816 | ## | |
817 | # Wait for the monitor to form quorum (optionally, of size N) | |
818 | # | |
819 | # @param timeout duration (lower-bound) to wait for quorum to be formed | |
820 | # @param quorumsize size of quorum to wait for | |
821 | # @return 0 on success, 1 on error | |
822 | # | |
823 | function wait_for_quorum() { | |
824 | local timeout=$1 | |
825 | local quorumsize=$2 | |
826 | ||
827 | if [[ -z "$timeout" ]]; then | |
828 | timeout=300 | |
829 | fi | |
830 | ||
831 | if [[ -z "$quorumsize" ]]; then | |
832 | timeout $timeout ceph mon_status --format=json >&/dev/null || return 1 | |
833 | return 0 | |
834 | fi | |
835 | ||
836 | no_quorum=1 | |
c07f9fc5 | 837 | wait_until=$((`date +%s` + $timeout)) |
7c673cae FG |
838 | while [[ $(date +%s) -lt $wait_until ]]; do |
839 | jqfilter='.quorum | length == '$quorumsize | |
840 | jqinput="$(timeout $timeout ceph mon_status --format=json 2>/dev/null)" | |
841 | res=$(echo $jqinput | jq "$jqfilter") | |
842 | if [[ "$res" == "true" ]]; then | |
843 | no_quorum=0 | |
844 | break | |
845 | fi | |
846 | done | |
847 | return $no_quorum | |
848 | } | |
849 | ||
850 | ####################################################################### | |
851 | ||
852 | ## | |
853 | # Return the PG of supporting the **objectname** stored in | |
854 | # **poolname**, as reported by ceph osd map. | |
855 | # | |
856 | # @param poolname an existing pool | |
857 | # @param objectname an objectname (may or may not exist) | |
858 | # @param STDOUT a PG | |
859 | # @return 0 on success, 1 on error | |
860 | # | |
861 | function get_pg() { | |
862 | local poolname=$1 | |
863 | local objectname=$2 | |
864 | ||
31f18b77 | 865 | ceph --format json osd map $poolname $objectname 2>/dev/null | jq -r '.pgid' |
7c673cae FG |
866 | } |
867 | ||
868 | function test_get_pg() { | |
869 | local dir=$1 | |
870 | ||
871 | setup $dir || return 1 | |
872 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
873 | run_mgr $dir x || return 1 | |
874 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 875 | create_rbd_pool || return 1 |
7c673cae FG |
876 | wait_for_clean || return 1 |
877 | get_pg rbd GROUP | grep --quiet '^[0-9]\.[0-9a-f][0-9a-f]*$' || return 1 | |
878 | teardown $dir || return 1 | |
879 | } | |
880 | ||
881 | ####################################################################### | |
882 | ||
883 | ## | |
884 | # Return the value of the **config**, obtained via the config get command | |
885 | # of the admin socket of **daemon**.**id**. | |
886 | # | |
887 | # @param daemon mon or osd | |
888 | # @param id mon or osd ID | |
889 | # @param config the configuration variable name as found in config_opts.h | |
890 | # @param STDOUT the config value | |
891 | # @return 0 on success, 1 on error | |
892 | # | |
893 | function get_config() { | |
894 | local daemon=$1 | |
895 | local id=$2 | |
896 | local config=$3 | |
897 | ||
898 | CEPH_ARGS='' \ | |
c07f9fc5 | 899 | ceph --format json daemon $(get_asok_path $daemon.$id) \ |
7c673cae | 900 | config get $config 2> /dev/null | \ |
31f18b77 | 901 | jq -r ".$config" |
7c673cae FG |
902 | } |
903 | ||
904 | function test_get_config() { | |
905 | local dir=$1 | |
906 | ||
907 | # override the default config using command line arg and check it | |
908 | setup $dir || return 1 | |
909 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
910 | test $(get_config mon a osd_pool_default_size) = 1 || return 1 | |
911 | run_mgr $dir x || return 1 | |
912 | run_osd $dir 0 --osd_max_scrubs=3 || return 1 | |
913 | test $(get_config osd 0 osd_max_scrubs) = 3 || return 1 | |
914 | teardown $dir || return 1 | |
915 | } | |
916 | ||
917 | ####################################################################### | |
918 | ||
919 | ## | |
920 | # Set the **config** to specified **value**, via the config set command | |
921 | # of the admin socket of **daemon**.**id** | |
922 | # | |
923 | # @param daemon mon or osd | |
924 | # @param id mon or osd ID | |
925 | # @param config the configuration variable name as found in config_opts.h | |
926 | # @param value the config value | |
927 | # @return 0 on success, 1 on error | |
928 | # | |
929 | function set_config() { | |
930 | local daemon=$1 | |
931 | local id=$2 | |
932 | local config=$3 | |
933 | local value=$4 | |
934 | ||
c07f9fc5 | 935 | test $(env CEPH_ARGS='' ceph --format json daemon $(get_asok_path $daemon.$id) \ |
31f18b77 FG |
936 | config set $config $value 2> /dev/null | \ |
937 | jq 'has("success")') == true | |
7c673cae FG |
938 | } |
939 | ||
940 | function test_set_config() { | |
941 | local dir=$1 | |
942 | ||
943 | setup $dir || return 1 | |
944 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
945 | test $(get_config mon a ms_crc_header) = true || return 1 | |
946 | set_config mon a ms_crc_header false || return 1 | |
947 | test $(get_config mon a ms_crc_header) = false || return 1 | |
948 | set_config mon a ms_crc_header true || return 1 | |
949 | test $(get_config mon a ms_crc_header) = true || return 1 | |
950 | teardown $dir || return 1 | |
951 | } | |
952 | ||
953 | ####################################################################### | |
954 | ||
955 | ## | |
956 | # Return the OSD id of the primary OSD supporting the **objectname** | |
957 | # stored in **poolname**, as reported by ceph osd map. | |
958 | # | |
959 | # @param poolname an existing pool | |
960 | # @param objectname an objectname (may or may not exist) | |
961 | # @param STDOUT the primary OSD id | |
962 | # @return 0 on success, 1 on error | |
963 | # | |
964 | function get_primary() { | |
965 | local poolname=$1 | |
966 | local objectname=$2 | |
967 | ||
31f18b77 FG |
968 | ceph --format json osd map $poolname $objectname 2>/dev/null | \ |
969 | jq '.acting_primary' | |
7c673cae FG |
970 | } |
971 | ||
972 | function test_get_primary() { | |
973 | local dir=$1 | |
974 | ||
975 | setup $dir || return 1 | |
976 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
977 | local osd=0 | |
978 | run_mgr $dir x || return 1 | |
979 | run_osd $dir $osd || return 1 | |
c07f9fc5 | 980 | create_rbd_pool || return 1 |
7c673cae FG |
981 | wait_for_clean || return 1 |
982 | test $(get_primary rbd GROUP) = $osd || return 1 | |
983 | teardown $dir || return 1 | |
984 | } | |
985 | ||
986 | ####################################################################### | |
987 | ||
988 | ## | |
989 | # Return the id of any OSD supporting the **objectname** stored in | |
990 | # **poolname**, as reported by ceph osd map, except the primary. | |
991 | # | |
992 | # @param poolname an existing pool | |
993 | # @param objectname an objectname (may or may not exist) | |
994 | # @param STDOUT the OSD id | |
995 | # @return 0 on success, 1 on error | |
996 | # | |
997 | function get_not_primary() { | |
998 | local poolname=$1 | |
999 | local objectname=$2 | |
1000 | ||
1001 | local primary=$(get_primary $poolname $objectname) | |
31f18b77 FG |
1002 | ceph --format json osd map $poolname $objectname 2>/dev/null | \ |
1003 | jq ".acting | map(select (. != $primary)) | .[0]" | |
7c673cae FG |
1004 | } |
1005 | ||
1006 | function test_get_not_primary() { | |
1007 | local dir=$1 | |
1008 | ||
1009 | setup $dir || return 1 | |
1010 | run_mon $dir a --osd_pool_default_size=2 || return 1 | |
1011 | run_mgr $dir x || return 1 | |
1012 | run_osd $dir 0 || return 1 | |
1013 | run_osd $dir 1 || return 1 | |
c07f9fc5 | 1014 | create_rbd_pool || return 1 |
7c673cae FG |
1015 | wait_for_clean || return 1 |
1016 | local primary=$(get_primary rbd GROUP) | |
1017 | local not_primary=$(get_not_primary rbd GROUP) | |
1018 | test $not_primary != $primary || return 1 | |
1019 | test $not_primary = 0 -o $not_primary = 1 || return 1 | |
1020 | teardown $dir || return 1 | |
1021 | } | |
1022 | ||
1023 | ####################################################################### | |
1024 | ||
1025 | ## | |
1026 | # Run ceph-objectstore-tool against the OSD **id** using the data path | |
1027 | # **dir**. The OSD is killed with TERM prior to running | |
1028 | # ceph-objectstore-tool because access to the data path is | |
1029 | # exclusive. The OSD is restarted after the command completes. The | |
1030 | # objectstore_tool returns after all PG are active+clean again. | |
1031 | # | |
1032 | # @param dir the data path of the OSD | |
1033 | # @param id the OSD id | |
1034 | # @param ... arguments to ceph-objectstore-tool | |
1035 | # @param STDIN the input of ceph-objectstore-tool | |
1036 | # @param STDOUT the output of ceph-objectstore-tool | |
1037 | # @return 0 on success, 1 on error | |
1038 | # | |
1039 | # The value of $ceph_osd_args will be passed to restarted osds | |
1040 | # | |
1041 | function objectstore_tool() { | |
1042 | local dir=$1 | |
1043 | shift | |
1044 | local id=$1 | |
1045 | shift | |
1046 | local osd_data=$dir/$id | |
1047 | ||
1048 | local osd_type=$(cat $osd_data/type) | |
1049 | ||
1050 | kill_daemons $dir TERM osd.$id >&2 < /dev/null || return 1 | |
1051 | ||
1052 | local journal_args | |
1053 | if [ "$objectstore_type" == "filestore" ]; then | |
1054 | journal_args=" --journal-path $osd_data/journal" | |
1055 | fi | |
1056 | ceph-objectstore-tool \ | |
7c673cae FG |
1057 | --data-path $osd_data \ |
1058 | $journal_args \ | |
1059 | "$@" || return 1 | |
1060 | activate_osd $dir $id $ceph_osd_args >&2 || return 1 | |
1061 | wait_for_clean >&2 | |
1062 | } | |
1063 | ||
1064 | function test_objectstore_tool() { | |
1065 | local dir=$1 | |
1066 | ||
1067 | setup $dir || return 1 | |
1068 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1069 | local osd=0 | |
1070 | run_mgr $dir x || return 1 | |
1071 | run_osd $dir $osd || return 1 | |
c07f9fc5 | 1072 | create_rbd_pool || return 1 |
7c673cae FG |
1073 | wait_for_clean || return 1 |
1074 | rados --pool rbd put GROUP /etc/group || return 1 | |
1075 | objectstore_tool $dir $osd GROUP get-bytes | \ | |
1076 | diff - /etc/group | |
1077 | ! objectstore_tool $dir $osd NOTEXISTS get-bytes || return 1 | |
1078 | teardown $dir || return 1 | |
1079 | } | |
1080 | ||
1081 | ####################################################################### | |
1082 | ||
1083 | ## | |
1084 | # Predicate checking if there is an ongoing recovery in the | |
1085 | # cluster. If any of the recovering_{keys,bytes,objects}_per_sec | |
1086 | # counters are reported by ceph status, it means recovery is in | |
1087 | # progress. | |
1088 | # | |
1089 | # @return 0 if recovery in progress, 1 otherwise | |
1090 | # | |
1091 | function get_is_making_recovery_progress() { | |
31f18b77 FG |
1092 | local recovery_progress |
1093 | recovery_progress+=".recovering_keys_per_sec + " | |
1094 | recovery_progress+=".recovering_bytes_per_sec + " | |
1095 | recovery_progress+=".recovering_objects_per_sec" | |
1096 | local progress=$(ceph --format json status 2>/dev/null | \ | |
1097 | jq -r ".pgmap | $recovery_progress") | |
1098 | test "$progress" != null | |
7c673cae FG |
1099 | } |
1100 | ||
1101 | function test_get_is_making_recovery_progress() { | |
1102 | local dir=$1 | |
1103 | ||
1104 | setup $dir || return 1 | |
1105 | run_mon $dir a || return 1 | |
1106 | run_mgr $dir x || return 1 | |
1107 | ! get_is_making_recovery_progress || return 1 | |
1108 | teardown $dir || return 1 | |
1109 | } | |
1110 | ||
1111 | ####################################################################### | |
1112 | ||
1113 | ## | |
1114 | # Return the number of active PGs in the cluster. A PG is active if | |
1115 | # ceph pg dump pgs reports it both **active** and **clean** and that | |
1116 | # not **stale**. | |
1117 | # | |
1118 | # @param STDOUT the number of active PGs | |
1119 | # @return 0 on success, 1 on error | |
1120 | # | |
1121 | function get_num_active_clean() { | |
31f18b77 FG |
1122 | local expression |
1123 | expression+="select(contains(\"active\") and contains(\"clean\")) | " | |
1124 | expression+="select(contains(\"stale\") | not)" | |
1125 | ceph --format json pg dump pgs 2>/dev/null | \ | |
1126 | jq "[.[] | .state | $expression] | length" | |
7c673cae FG |
1127 | } |
1128 | ||
1129 | function test_get_num_active_clean() { | |
1130 | local dir=$1 | |
1131 | ||
1132 | setup $dir || return 1 | |
1133 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1134 | run_mgr $dir x || return 1 | |
1135 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1136 | create_rbd_pool || return 1 |
7c673cae FG |
1137 | wait_for_clean || return 1 |
1138 | local num_active_clean=$(get_num_active_clean) | |
1139 | test "$num_active_clean" = $PG_NUM || return 1 | |
1140 | teardown $dir || return 1 | |
1141 | } | |
1142 | ||
1143 | ####################################################################### | |
1144 | ||
1145 | ## | |
1146 | # Return the number of PGs in the cluster, according to | |
1147 | # ceph pg dump pgs. | |
1148 | # | |
1149 | # @param STDOUT the number of PGs | |
1150 | # @return 0 on success, 1 on error | |
1151 | # | |
1152 | function get_num_pgs() { | |
31f18b77 | 1153 | ceph --format json status 2>/dev/null | jq '.pgmap.num_pgs' |
7c673cae FG |
1154 | } |
1155 | ||
1156 | function test_get_num_pgs() { | |
1157 | local dir=$1 | |
1158 | ||
1159 | setup $dir || return 1 | |
1160 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1161 | run_mgr $dir x || return 1 | |
1162 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1163 | create_rbd_pool || return 1 |
7c673cae FG |
1164 | wait_for_clean || return 1 |
1165 | local num_pgs=$(get_num_pgs) | |
1166 | test "$num_pgs" -gt 0 || return 1 | |
1167 | teardown $dir || return 1 | |
1168 | } | |
1169 | ||
1170 | ####################################################################### | |
1171 | ||
c07f9fc5 FG |
1172 | ## |
1173 | # Return the OSD ids in use by at least one PG in the cluster (either | |
1174 | # in the up or the acting set), according to ceph pg dump pgs. Every | |
1175 | # OSD id shows as many times as they are used in up and acting sets. | |
1176 | # If an OSD id is in both the up and acting set of a given PG, it will | |
1177 | # show twice. | |
1178 | # | |
1179 | # @param STDOUT a sorted list of OSD ids | |
1180 | # @return 0 on success, 1 on error | |
1181 | # | |
1182 | function get_osd_id_used_by_pgs() { | |
1183 | ceph --format json pg dump pgs 2>/dev/null | jq '.[] | .up[], .acting[]' | sort | |
1184 | } | |
1185 | ||
1186 | function test_get_osd_id_used_by_pgs() { | |
1187 | local dir=$1 | |
1188 | ||
1189 | setup $dir || return 1 | |
1190 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1191 | run_mgr $dir x || return 1 | |
1192 | run_osd $dir 0 || return 1 | |
1193 | create_rbd_pool || return 1 | |
1194 | wait_for_clean || return 1 | |
1195 | local osd_ids=$(get_osd_id_used_by_pgs | uniq) | |
1196 | test "$osd_ids" = "0" || return 1 | |
1197 | teardown $dir || return 1 | |
1198 | } | |
1199 | ||
1200 | ####################################################################### | |
1201 | ||
1202 | ## | |
1203 | # Wait until the OSD **id** shows **count** times in the | |
1204 | # PGs (see get_osd_id_used_by_pgs for more information about | |
1205 | # how OSD ids are counted). | |
1206 | # | |
1207 | # @param id the OSD id | |
1208 | # @param count the number of time it must show in the PGs | |
1209 | # @return 0 on success, 1 on error | |
1210 | # | |
1211 | function wait_osd_id_used_by_pgs() { | |
1212 | local id=$1 | |
1213 | local count=$2 | |
1214 | ||
1215 | status=1 | |
1216 | for ((i=0; i < $TIMEOUT / 5; i++)); do | |
1217 | echo $i | |
1218 | if ! test $(get_osd_id_used_by_pgs | grep -c $id) = $count ; then | |
1219 | sleep 5 | |
1220 | else | |
1221 | status=0 | |
1222 | break | |
1223 | fi | |
1224 | done | |
1225 | return $status | |
1226 | } | |
1227 | ||
1228 | function test_wait_osd_id_used_by_pgs() { | |
1229 | local dir=$1 | |
1230 | ||
1231 | setup $dir || return 1 | |
1232 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1233 | run_mgr $dir x || return 1 | |
1234 | run_osd $dir 0 || return 1 | |
1235 | create_rbd_pool || return 1 | |
1236 | wait_for_clean || return 1 | |
1237 | wait_osd_id_used_by_pgs 0 8 || return 1 | |
1238 | ! TIMEOUT=1 wait_osd_id_used_by_pgs 123 5 || return 1 | |
1239 | teardown $dir || return 1 | |
1240 | } | |
1241 | ||
1242 | ####################################################################### | |
1243 | ||
7c673cae FG |
1244 | ## |
1245 | # Return the date and time of the last completed scrub for **pgid**, | |
1246 | # as reported by ceph pg dump pgs. Note that a repair also sets this | |
1247 | # date. | |
1248 | # | |
1249 | # @param pgid the id of the PG | |
1250 | # @param STDOUT the date and time of the last scrub | |
1251 | # @return 0 on success, 1 on error | |
1252 | # | |
1253 | function get_last_scrub_stamp() { | |
1254 | local pgid=$1 | |
1255 | local sname=${2:-last_scrub_stamp} | |
31f18b77 FG |
1256 | ceph --format json pg dump pgs 2>/dev/null | \ |
1257 | jq -r ".[] | select(.pgid==\"$pgid\") | .$sname" | |
7c673cae FG |
1258 | } |
1259 | ||
1260 | function test_get_last_scrub_stamp() { | |
1261 | local dir=$1 | |
1262 | ||
1263 | setup $dir || return 1 | |
1264 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1265 | run_mgr $dir x || return 1 | |
1266 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1267 | create_rbd_pool || return 1 |
7c673cae | 1268 | wait_for_clean || return 1 |
224ce89b | 1269 | stamp=$(get_last_scrub_stamp 2.0) |
7c673cae FG |
1270 | test -n "$stamp" || return 1 |
1271 | teardown $dir || return 1 | |
1272 | } | |
1273 | ||
1274 | ####################################################################### | |
1275 | ||
1276 | ## | |
1277 | # Predicate checking if the cluster is clean, i.e. all of its PGs are | |
1278 | # in a clean state (see get_num_active_clean for a definition). | |
1279 | # | |
1280 | # @return 0 if the cluster is clean, 1 otherwise | |
1281 | # | |
1282 | function is_clean() { | |
1283 | num_pgs=$(get_num_pgs) | |
1284 | test $num_pgs != 0 || return 1 | |
1285 | test $(get_num_active_clean) = $num_pgs || return 1 | |
1286 | } | |
1287 | ||
1288 | function test_is_clean() { | |
1289 | local dir=$1 | |
1290 | ||
1291 | setup $dir || return 1 | |
1292 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1293 | run_mgr $dir x || return 1 | |
1294 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1295 | create_rbd_pool || return 1 |
7c673cae FG |
1296 | wait_for_clean || return 1 |
1297 | is_clean || return 1 | |
1298 | teardown $dir || return 1 | |
1299 | } | |
1300 | ||
1301 | ####################################################################### | |
1302 | ||
1303 | ## | |
1304 | # Return a list of numbers that are increasingly larger and whose | |
1305 | # total is **timeout** seconds. It can be used to have short sleep | |
1306 | # delay while waiting for an event on a fast machine. But if running | |
1307 | # very slowly the larger delays avoid stressing the machine even | |
1308 | # further or spamming the logs. | |
1309 | # | |
1310 | # @param timeout sum of all delays, in seconds | |
1311 | # @return a list of sleep delays | |
1312 | # | |
1313 | function get_timeout_delays() { | |
1314 | local trace=$(shopt -q -o xtrace && echo true || echo false) | |
1315 | $trace && shopt -u -o xtrace | |
1316 | local timeout=$1 | |
1317 | local first_step=${2:-1} | |
1318 | ||
1319 | local i | |
1320 | local total="0" | |
1321 | i=$first_step | |
1322 | while test "$(echo $total + $i \<= $timeout | bc -l)" = "1"; do | |
1323 | echo -n "$i " | |
1324 | total=$(echo $total + $i | bc -l) | |
1325 | i=$(echo $i \* 2 | bc -l) | |
1326 | done | |
1327 | if test "$(echo $total \< $timeout | bc -l)" = "1"; then | |
1328 | echo -n $(echo $timeout - $total | bc -l) | |
1329 | fi | |
1330 | $trace && shopt -s -o xtrace | |
1331 | } | |
1332 | ||
1333 | function test_get_timeout_delays() { | |
1334 | test "$(get_timeout_delays 1)" = "1 " || return 1 | |
1335 | test "$(get_timeout_delays 5)" = "1 2 2" || return 1 | |
1336 | test "$(get_timeout_delays 6)" = "1 2 3" || return 1 | |
1337 | test "$(get_timeout_delays 7)" = "1 2 4 " || return 1 | |
1338 | test "$(get_timeout_delays 8)" = "1 2 4 1" || return 1 | |
1339 | test "$(get_timeout_delays 1 .1)" = ".1 .2 .4 .3" || return 1 | |
1340 | test "$(get_timeout_delays 1.5 .1)" = ".1 .2 .4 .8 " || return 1 | |
1341 | test "$(get_timeout_delays 5 .1)" = ".1 .2 .4 .8 1.6 1.9" || return 1 | |
1342 | test "$(get_timeout_delays 6 .1)" = ".1 .2 .4 .8 1.6 2.9" || return 1 | |
1343 | test "$(get_timeout_delays 6.3 .1)" = ".1 .2 .4 .8 1.6 3.2 " || return 1 | |
1344 | test "$(get_timeout_delays 20 .1)" = ".1 .2 .4 .8 1.6 3.2 6.4 7.3" || return 1 | |
1345 | } | |
1346 | ||
1347 | ####################################################################### | |
1348 | ||
1349 | ## | |
1350 | # Wait until the cluster becomes clean or if it does not make progress | |
1351 | # for $TIMEOUT seconds. | |
1352 | # Progress is measured either via the **get_is_making_recovery_progress** | |
1353 | # predicate or if the number of clean PGs changes (as returned by get_num_active_clean) | |
1354 | # | |
1355 | # @return 0 if the cluster is clean, 1 otherwise | |
1356 | # | |
1357 | function wait_for_clean() { | |
1358 | local num_active_clean=-1 | |
1359 | local cur_active_clean | |
1360 | local -a delays=($(get_timeout_delays $TIMEOUT .1)) | |
1361 | local -i loop=0 | |
31f18b77 FG |
1362 | |
1363 | while test $(get_num_pgs) == 0 ; do | |
1364 | sleep 1 | |
1365 | done | |
7c673cae FG |
1366 | |
1367 | while true ; do | |
1368 | # Comparing get_num_active_clean & get_num_pgs is used to determine | |
1369 | # if the cluster is clean. That's almost an inline of is_clean() to | |
1370 | # get more performance by avoiding multiple calls of get_num_active_clean. | |
1371 | cur_active_clean=$(get_num_active_clean) | |
1372 | test $cur_active_clean = $(get_num_pgs) && break | |
1373 | if test $cur_active_clean != $num_active_clean ; then | |
1374 | loop=0 | |
1375 | num_active_clean=$cur_active_clean | |
1376 | elif get_is_making_recovery_progress ; then | |
1377 | loop=0 | |
1378 | elif (( $loop >= ${#delays[*]} )) ; then | |
1379 | ceph report | |
1380 | return 1 | |
1381 | fi | |
1382 | sleep ${delays[$loop]} | |
1383 | loop+=1 | |
1384 | done | |
1385 | return 0 | |
1386 | } | |
1387 | ||
1388 | function test_wait_for_clean() { | |
1389 | local dir=$1 | |
1390 | ||
1391 | setup $dir || return 1 | |
1392 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1393 | run_mgr $dir x || return 1 | |
c07f9fc5 | 1394 | create_rbd_pool || return 1 |
7c673cae FG |
1395 | ! TIMEOUT=1 wait_for_clean || return 1 |
1396 | run_osd $dir 0 || return 1 | |
1397 | wait_for_clean || return 1 | |
1398 | teardown $dir || return 1 | |
1399 | } | |
1400 | ||
1401 | ####################################################################### | |
1402 | ||
1403 | ## | |
1404 | # Wait until the cluster becomes HEALTH_OK again or if it does not make progress | |
1405 | # for $TIMEOUT seconds. | |
1406 | # | |
1407 | # @return 0 if the cluster is HEALTHY, 1 otherwise | |
1408 | # | |
1409 | function wait_for_health() { | |
1410 | local grepstr=$1 | |
1411 | local -a delays=($(get_timeout_delays $TIMEOUT .1)) | |
1412 | local -i loop=0 | |
1413 | ||
1414 | while ! ceph health detail | grep "$grepstr" ; do | |
1415 | if (( $loop >= ${#delays[*]} )) ; then | |
1416 | ceph health detail | |
1417 | return 1 | |
1418 | fi | |
1419 | sleep ${delays[$loop]} | |
1420 | loop+=1 | |
1421 | done | |
1422 | } | |
1423 | ||
1424 | function wait_for_health_ok() { | |
1425 | wait_for_health "HEALTH_OK" || return 1 | |
1426 | } | |
1427 | ||
1428 | function test_wait_for_health_ok() { | |
1429 | local dir=$1 | |
1430 | ||
1431 | setup $dir || return 1 | |
1432 | run_mon $dir a --osd_pool_default_size=1 --osd_failsafe_full_ratio=.99 --mon_pg_warn_min_per_osd=0 || return 1 | |
31f18b77 | 1433 | run_mgr $dir x --mon_pg_warn_min_per_osd=0 || return 1 |
7c673cae | 1434 | run_osd $dir 0 || return 1 |
224ce89b WB |
1435 | kill_daemons $dir TERM osd || return 1 |
1436 | ! TIMEOUT=1 wait_for_health_ok || return 1 | |
1437 | activate_osd $dir 0 || return 1 | |
7c673cae FG |
1438 | wait_for_health_ok || return 1 |
1439 | teardown $dir || return 1 | |
1440 | } | |
1441 | ||
1442 | ||
1443 | ####################################################################### | |
1444 | ||
1445 | ## | |
1446 | # Run repair on **pgid** and wait until it completes. The repair | |
1447 | # function will fail if repair does not complete within $TIMEOUT | |
1448 | # seconds. | |
1449 | # | |
1450 | # @param pgid the id of the PG | |
1451 | # @return 0 on success, 1 on error | |
1452 | # | |
1453 | function repair() { | |
1454 | local pgid=$1 | |
1455 | local last_scrub=$(get_last_scrub_stamp $pgid) | |
1456 | ceph pg repair $pgid | |
1457 | wait_for_scrub $pgid "$last_scrub" | |
1458 | } | |
1459 | ||
1460 | function test_repair() { | |
1461 | local dir=$1 | |
1462 | ||
1463 | setup $dir || return 1 | |
1464 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1465 | run_mgr $dir x || return 1 | |
1466 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1467 | create_rbd_pool || return 1 |
7c673cae | 1468 | wait_for_clean || return 1 |
224ce89b | 1469 | repair 2.0 || return 1 |
7c673cae | 1470 | kill_daemons $dir KILL osd || return 1 |
224ce89b | 1471 | ! TIMEOUT=1 repair 2.0 || return 1 |
7c673cae FG |
1472 | teardown $dir || return 1 |
1473 | } | |
1474 | ####################################################################### | |
1475 | ||
1476 | ## | |
1477 | # Run scrub on **pgid** and wait until it completes. The pg_scrub | |
1478 | # function will fail if repair does not complete within $TIMEOUT | |
1479 | # seconds. The pg_scrub is complete whenever the | |
1480 | # **get_last_scrub_stamp** function reports a timestamp different from | |
1481 | # the one stored before starting the scrub. | |
1482 | # | |
1483 | # @param pgid the id of the PG | |
1484 | # @return 0 on success, 1 on error | |
1485 | # | |
1486 | function pg_scrub() { | |
1487 | local pgid=$1 | |
1488 | local last_scrub=$(get_last_scrub_stamp $pgid) | |
1489 | ceph pg scrub $pgid | |
1490 | wait_for_scrub $pgid "$last_scrub" | |
1491 | } | |
1492 | ||
1493 | function pg_deep_scrub() { | |
1494 | local pgid=$1 | |
1495 | local last_scrub=$(get_last_scrub_stamp $pgid last_deep_scrub_stamp) | |
1496 | ceph pg deep-scrub $pgid | |
1497 | wait_for_scrub $pgid "$last_scrub" last_deep_scrub_stamp | |
1498 | } | |
1499 | ||
1500 | function test_pg_scrub() { | |
1501 | local dir=$1 | |
1502 | ||
1503 | setup $dir || return 1 | |
1504 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1505 | run_mgr $dir x || return 1 | |
1506 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1507 | create_rbd_pool || return 1 |
7c673cae | 1508 | wait_for_clean || return 1 |
224ce89b | 1509 | pg_scrub 2.0 || return 1 |
7c673cae | 1510 | kill_daemons $dir KILL osd || return 1 |
224ce89b | 1511 | ! TIMEOUT=1 pg_scrub 2.0 || return 1 |
7c673cae FG |
1512 | teardown $dir || return 1 |
1513 | } | |
1514 | ||
1515 | ####################################################################### | |
1516 | ||
1517 | ## | |
1518 | # Run the *command* and expect it to fail (i.e. return a non zero status). | |
1519 | # The output (stderr and stdout) is stored in a temporary file in *dir* | |
1520 | # and is expected to contain the string *expected*. | |
1521 | # | |
1522 | # Return 0 if the command failed and the string was found. Otherwise | |
1523 | # return 1 and cat the full output of the command on stderr for debug. | |
1524 | # | |
1525 | # @param dir temporary directory to store the output | |
1526 | # @param expected string to look for in the output | |
1527 | # @param command ... the command and its arguments | |
1528 | # @return 0 on success, 1 on error | |
1529 | # | |
1530 | ||
1531 | function expect_failure() { | |
1532 | local dir=$1 | |
1533 | shift | |
1534 | local expected="$1" | |
1535 | shift | |
1536 | local success | |
1537 | ||
1538 | if "$@" > $dir/out 2>&1 ; then | |
1539 | success=true | |
1540 | else | |
1541 | success=false | |
1542 | fi | |
1543 | ||
1544 | if $success || ! grep --quiet "$expected" $dir/out ; then | |
1545 | cat $dir/out >&2 | |
1546 | return 1 | |
1547 | else | |
1548 | return 0 | |
1549 | fi | |
1550 | } | |
1551 | ||
1552 | function test_expect_failure() { | |
1553 | local dir=$1 | |
1554 | ||
1555 | setup $dir || return 1 | |
1556 | expect_failure $dir FAIL bash -c 'echo FAIL ; exit 1' || return 1 | |
1557 | # the command did not fail | |
1558 | ! expect_failure $dir FAIL bash -c 'echo FAIL ; exit 0' > $dir/out || return 1 | |
1559 | grep --quiet FAIL $dir/out || return 1 | |
1560 | # the command failed but the output does not contain the expected string | |
1561 | ! expect_failure $dir FAIL bash -c 'echo UNEXPECTED ; exit 1' > $dir/out || return 1 | |
1562 | ! grep --quiet FAIL $dir/out || return 1 | |
1563 | teardown $dir || return 1 | |
1564 | } | |
1565 | ||
1566 | ####################################################################### | |
1567 | ||
1568 | ## | |
1569 | # Given the *last_scrub*, wait for scrub to happen on **pgid**. It | |
1570 | # will fail if scrub does not complete within $TIMEOUT seconds. The | |
1571 | # repair is complete whenever the **get_last_scrub_stamp** function | |
1572 | # reports a timestamp different from the one given in argument. | |
1573 | # | |
1574 | # @param pgid the id of the PG | |
1575 | # @param last_scrub timestamp of the last scrub for *pgid* | |
1576 | # @return 0 on success, 1 on error | |
1577 | # | |
1578 | function wait_for_scrub() { | |
1579 | local pgid=$1 | |
1580 | local last_scrub="$2" | |
1581 | local sname=${3:-last_scrub_stamp} | |
1582 | ||
1583 | for ((i=0; i < $TIMEOUT; i++)); do | |
1584 | if test "$last_scrub" != "$(get_last_scrub_stamp $pgid $sname)" ; then | |
1585 | return 0 | |
1586 | fi | |
1587 | sleep 1 | |
1588 | done | |
1589 | return 1 | |
1590 | } | |
1591 | ||
1592 | function test_wait_for_scrub() { | |
1593 | local dir=$1 | |
1594 | ||
1595 | setup $dir || return 1 | |
1596 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1597 | run_mgr $dir x || return 1 | |
1598 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1599 | create_rbd_pool || return 1 |
7c673cae | 1600 | wait_for_clean || return 1 |
224ce89b | 1601 | local pgid=2.0 |
7c673cae FG |
1602 | ceph pg repair $pgid |
1603 | local last_scrub=$(get_last_scrub_stamp $pgid) | |
1604 | wait_for_scrub $pgid "$last_scrub" || return 1 | |
1605 | kill_daemons $dir KILL osd || return 1 | |
1606 | last_scrub=$(get_last_scrub_stamp $pgid) | |
1607 | ! TIMEOUT=1 wait_for_scrub $pgid "$last_scrub" || return 1 | |
1608 | teardown $dir || return 1 | |
1609 | } | |
1610 | ||
1611 | ####################################################################### | |
1612 | ||
1613 | ## | |
1614 | # Return 0 if the erasure code *plugin* is available, 1 otherwise. | |
1615 | # | |
1616 | # @param plugin erasure code plugin | |
1617 | # @return 0 on success, 1 on error | |
1618 | # | |
1619 | ||
1620 | function erasure_code_plugin_exists() { | |
1621 | local plugin=$1 | |
1622 | local status | |
1623 | local grepstr | |
1624 | local s | |
1625 | case `uname` in | |
1626 | FreeBSD) grepstr="Cannot open.*$plugin" ;; | |
1627 | *) grepstr="$plugin.*No such file" ;; | |
1628 | esac | |
1629 | ||
1630 | s=$(ceph osd erasure-code-profile set TESTPROFILE plugin=$plugin 2>&1) | |
1631 | local status=$? | |
1632 | if [ $status -eq 0 ]; then | |
1633 | ceph osd erasure-code-profile rm TESTPROFILE | |
1634 | elif ! echo $s | grep --quiet "$grepstr" ; then | |
1635 | status=1 | |
1636 | # display why the string was rejected. | |
1637 | echo $s | |
1638 | fi | |
1639 | return $status | |
1640 | } | |
1641 | ||
1642 | function test_erasure_code_plugin_exists() { | |
1643 | local dir=$1 | |
1644 | ||
1645 | setup $dir || return 1 | |
1646 | run_mon $dir a || return 1 | |
1647 | run_mgr $dir x || return 1 | |
1648 | erasure_code_plugin_exists jerasure || return 1 | |
1649 | ! erasure_code_plugin_exists FAKE || return 1 | |
1650 | teardown $dir || return 1 | |
1651 | } | |
1652 | ||
1653 | ####################################################################### | |
1654 | ||
1655 | ## | |
1656 | # Display all log files from **dir** on stdout. | |
1657 | # | |
1658 | # @param dir directory in which all data is stored | |
1659 | # | |
1660 | ||
1661 | function display_logs() { | |
1662 | local dir=$1 | |
1663 | ||
1664 | find $dir -maxdepth 1 -name '*.log' | \ | |
1665 | while read file ; do | |
1666 | echo "======================= $file" | |
1667 | cat $file | |
1668 | done | |
1669 | } | |
1670 | ||
1671 | function test_display_logs() { | |
1672 | local dir=$1 | |
1673 | ||
1674 | setup $dir || return 1 | |
1675 | run_mon $dir a || return 1 | |
1676 | kill_daemons $dir || return 1 | |
1677 | display_logs $dir > $dir/log.out | |
1678 | grep --quiet mon.a.log $dir/log.out || return 1 | |
1679 | teardown $dir || return 1 | |
1680 | } | |
1681 | ||
1682 | ####################################################################### | |
1683 | ## | |
1684 | # Spawn a command in background and save the pid in the variable name | |
1685 | # passed in argument. To make the output reading easier, the output is | |
1686 | # prepend with the process id. | |
1687 | # | |
1688 | # Example: | |
1689 | # pids1="" | |
1690 | # run_in_background pids1 bash -c 'sleep 1; exit 1' | |
1691 | # | |
1692 | # @param pid_variable the variable name (not value) where the pids will be stored | |
1693 | # @param ... the command to execute | |
1694 | # @return only the pid_variable output should be considered and used with **wait_background** | |
1695 | # | |
1696 | function run_in_background() { | |
1697 | local pid_variable=$1 | |
1698 | shift; | |
1699 | # Execute the command and prepend the output with its pid | |
1700 | # We enforce to return the exit status of the command and not the awk one. | |
1701 | ("$@" |& awk '{ a[i++] = $0 }END{for (i = 0; i in a; ++i) { print "'$$': " a[i]} }'; return ${PIPESTATUS[0]}) >&2 & | |
1702 | eval "$pid_variable+=\" $!\"" | |
1703 | } | |
1704 | ||
1705 | function test_run_in_background() { | |
1706 | local pids | |
1707 | run_in_background pids sleep 1 | |
1708 | run_in_background pids sleep 1 | |
1709 | test $(echo $pids | wc -w) = 2 || return 1 | |
1710 | wait $pids || return 1 | |
1711 | } | |
1712 | ||
1713 | ####################################################################### | |
1714 | ## | |
1715 | # Wait for pids running in background to complete. | |
1716 | # This function is usually used after a **run_in_background** call | |
1717 | # Example: | |
1718 | # pids1="" | |
1719 | # run_in_background pids1 bash -c 'sleep 1; exit 1' | |
1720 | # wait_background pids1 | |
1721 | # | |
1722 | # @param pids The variable name that contains the active PIDS. Set as empty at then end of the function. | |
1723 | # @return returns 1 if at least one process exits in error unless returns 0 | |
1724 | # | |
1725 | function wait_background() { | |
1726 | # We extract the PIDS from the variable name | |
1727 | pids=${!1} | |
1728 | ||
1729 | return_code=0 | |
1730 | for pid in $pids; do | |
1731 | if ! wait $pid; then | |
1732 | # If one process failed then return 1 | |
1733 | return_code=1 | |
1734 | fi | |
1735 | done | |
1736 | ||
1737 | # We empty the variable reporting that all process ended | |
1738 | eval "$1=''" | |
1739 | ||
1740 | return $return_code | |
1741 | } | |
1742 | ||
1743 | ||
1744 | function test_wait_background() { | |
1745 | local pids="" | |
1746 | run_in_background pids bash -c "sleep 1; exit 1" | |
1747 | run_in_background pids bash -c "sleep 2; exit 0" | |
1748 | wait_background pids | |
1749 | if [ $? -ne 1 ]; then return 1; fi | |
1750 | ||
1751 | run_in_background pids bash -c "sleep 1; exit 0" | |
1752 | run_in_background pids bash -c "sleep 2; exit 0" | |
1753 | wait_background pids | |
1754 | if [ $? -ne 0 ]; then return 1; fi | |
1755 | ||
1756 | if [ ! -z "$pids" ]; then return 1; fi | |
1757 | } | |
1758 | ||
31f18b77 FG |
1759 | function flush_pg_stats() |
1760 | { | |
1761 | local timeout=${1:-$TIMEOUT} | |
1762 | ||
1763 | ids=`ceph osd ls` | |
1764 | seqs='' | |
1765 | for osd in $ids; do | |
1766 | seq=`ceph tell osd.$osd flush_pg_stats` | |
1767 | seqs="$seqs $osd-$seq" | |
1768 | done | |
1769 | ||
1770 | for s in $seqs; do | |
1771 | osd=`echo $s | cut -d - -f 1` | |
1772 | seq=`echo $s | cut -d - -f 2` | |
1773 | echo "waiting osd.$osd seq $seq" | |
1774 | while test $(ceph osd last-stat-seq $osd) -lt $seq; do | |
1775 | sleep 1 | |
1776 | if [ $((timeout--)) -eq 0 ]; then | |
1777 | return 1 | |
1778 | fi | |
1779 | done | |
1780 | done | |
1781 | } | |
1782 | ||
1783 | function test_flush_pg_stats() | |
1784 | { | |
1785 | local dir=$1 | |
1786 | ||
1787 | setup $dir || return 1 | |
1788 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1789 | run_mgr $dir x || return 1 | |
1790 | run_osd $dir 0 || return 1 | |
c07f9fc5 | 1791 | create_rbd_pool || return 1 |
31f18b77 FG |
1792 | rados -p rbd put obj /etc/group |
1793 | flush_pg_stats | |
1794 | local jq_filter='.pools | .[] | select(.name == "rbd") | .stats' | |
1795 | raw_bytes_used=`ceph df detail --format=json | jq "$jq_filter.raw_bytes_used"` | |
1796 | bytes_used=`ceph df detail --format=json | jq "$jq_filter.bytes_used"` | |
1797 | test $raw_bytes_used > 0 || return 1 | |
1798 | test $raw_bytes_used == $bytes_used || return 1 | |
1799 | } | |
1800 | ||
7c673cae FG |
1801 | ####################################################################### |
1802 | ||
1803 | ## | |
1804 | # Call the **run** function (which must be defined by the caller) with | |
1805 | # the **dir** argument followed by the caller argument list. | |
1806 | # | |
1807 | # If the **run** function returns on error, all logs found in **dir** | |
1808 | # are displayed for diagnostic purposes. | |
1809 | # | |
1810 | # **teardown** function is called when the **run** function returns | |
1811 | # (on success or on error), to cleanup leftovers. The CEPH_CONF is set | |
1812 | # to /dev/null and CEPH_ARGS is unset so that the tests are protected from | |
1813 | # external interferences. | |
1814 | # | |
1815 | # It is the responsibility of the **run** function to call the | |
1816 | # **setup** function to prepare the test environment (create a temporary | |
1817 | # directory etc.). | |
1818 | # | |
1819 | # The shell is required (via PS4) to display the function and line | |
1820 | # number whenever a statement is executed to help debugging. | |
1821 | # | |
1822 | # @param dir directory in which all data is stored | |
1823 | # @param ... arguments passed transparently to **run** | |
1824 | # @return 0 on success, 1 on error | |
1825 | # | |
1826 | function main() { | |
1827 | local dir=td/$1 | |
1828 | shift | |
1829 | ||
1830 | shopt -s -o xtrace | |
1831 | PS4='${BASH_SOURCE[0]}:$LINENO: ${FUNCNAME[0]}: ' | |
1832 | ||
1833 | export PATH=${CEPH_BUILD_VIRTUALENV}/ceph-disk-virtualenv/bin:${CEPH_BUILD_VIRTUALENV}/ceph-detect-init-virtualenv/bin:.:$PATH # make sure program from sources are preferred | |
1834 | #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 | |
1835 | ||
1836 | export CEPH_CONF=/dev/null | |
1837 | unset CEPH_ARGS | |
1838 | ||
1839 | local code | |
1840 | if run $dir "$@" ; then | |
1841 | code=0 | |
1842 | else | |
1843 | display_logs $dir | |
1844 | code=1 | |
1845 | fi | |
1846 | teardown $dir || return 1 | |
1847 | return $code | |
1848 | } | |
1849 | ||
1850 | ####################################################################### | |
1851 | ||
1852 | function run_tests() { | |
1853 | shopt -s -o xtrace | |
1854 | PS4='${BASH_SOURCE[0]}:$LINENO: ${FUNCNAME[0]}: ' | |
1855 | ||
1856 | export PATH=${CEPH_BUILD_VIRTUALENV}/ceph-disk-virtualenv/bin:${CEPH_BUILD_VIRTUALENV}/ceph-detect-init-virtualenv/bin:.:$PATH # make sure program from sources are preferred | |
1857 | #export PATH=$CEPH_ROOT/src/ceph-disk/virtualenv/bin:$CEPH_ROOT/src/ceph-detect-init/virtualenv/bin:.:$PATH # make sure program from sources are preferred | |
1858 | ||
1859 | export CEPH_MON="127.0.0.1:7109" # git grep '\<7109\>' : there must be only one | |
1860 | export CEPH_ARGS | |
1861 | CEPH_ARGS+="--fsid=$(uuidgen) --auth-supported=none " | |
1862 | CEPH_ARGS+="--mon-host=$CEPH_MON " | |
1863 | export CEPH_CONF=/dev/null | |
1864 | ||
1865 | local funcs=${@:-$(set | sed -n -e 's/^\(test_[0-9a-z_]*\) .*/\1/p')} | |
1866 | local dir=td/ceph-helpers | |
1867 | ||
1868 | for func in $funcs ; do | |
1869 | $func $dir || return 1 | |
1870 | done | |
1871 | } | |
1872 | ||
1873 | if test "$1" = TESTS ; then | |
1874 | shift | |
1875 | run_tests "$@" | |
1876 | fi | |
1877 | ||
224ce89b WB |
1878 | # NOTE: |
1879 | # jq only support --exit-status|-e from version 1.4 forwards, which makes | |
1880 | # returning on error waaaay prettier and straightforward. | |
1881 | # However, the current automated upstream build is running with v1.3, | |
1882 | # which has no idea what -e is. Hence the convoluted error checking we | |
1883 | # need. Sad. | |
1884 | # The next time someone changes this code, please check if v1.4 is now | |
1885 | # a thing, and, if so, please change these to use -e. Thanks. | |
1886 | ||
1887 | # jq '.all.supported | select([.[] == "foo"] | any)' | |
1888 | function jq_success() { | |
1889 | input="$1" | |
1890 | filter="$2" | |
1891 | expects="\"$3\"" | |
1892 | ||
1893 | in_escaped=$(printf %s "$input" | sed "s/'/'\\\\''/g") | |
1894 | filter_escaped=$(printf %s "$filter" | sed "s/'/'\\\\''/g") | |
1895 | ||
1896 | ret=$(echo "$in_escaped" | jq "$filter_escaped") | |
1897 | if [[ "$ret" == "true" ]]; then | |
1898 | return 0 | |
1899 | elif [[ -n "$expects" ]]; then | |
1900 | if [[ "$ret" == "$expects" ]]; then | |
1901 | return 0 | |
1902 | fi | |
1903 | fi | |
1904 | return 1 | |
1905 | input=$1 | |
1906 | filter=$2 | |
1907 | expects="$3" | |
1908 | ||
1909 | ret="$(echo $input | jq \"$filter\")" | |
1910 | if [[ "$ret" == "true" ]]; then | |
1911 | return 0 | |
1912 | elif [[ -n "$expects" && "$ret" == "$expects" ]]; then | |
1913 | return 0 | |
1914 | fi | |
1915 | return 1 | |
1916 | } | |
1917 | ||
7c673cae | 1918 | # Local Variables: |
c07f9fc5 | 1919 | # compile-command: "cd ../../src ; make -j4 && ../qa/standalone/ceph-helpers.sh TESTS # test_get_config" |
7c673cae | 1920 | # End: |