]>
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 | |
32 | if [ `uname` = FreeBSD ]; then | |
33 | SED=gsed | |
34 | else | |
35 | SED=sed | |
36 | fi | |
37 | ||
38 | #! @file ceph-helpers.sh | |
39 | # @brief Toolbox to manage Ceph cluster dedicated to testing | |
40 | # | |
41 | # Example use case: | |
42 | # | |
43 | # ~~~~~~~~~~~~~~~~{.sh} | |
44 | # source ceph-helpers.sh | |
45 | # | |
46 | # function mytest() { | |
47 | # # cleanup leftovers and reset mydir | |
48 | # setup mydir | |
49 | # # create a cluster with one monitor and three osds | |
50 | # run_mon mydir a | |
51 | # run_osd mydir 0 | |
52 | # run_osd mydir 2 | |
53 | # run_osd mydir 3 | |
54 | # # put and get an object | |
55 | # rados --pool rbd put GROUP /etc/group | |
56 | # rados --pool rbd get GROUP /tmp/GROUP | |
57 | # # stop the cluster and cleanup the directory | |
58 | # teardown mydir | |
59 | # } | |
60 | # ~~~~~~~~~~~~~~~~ | |
61 | # | |
62 | # The focus is on simplicity and efficiency, in the context of | |
63 | # functional tests. The output is intentionally very verbose | |
64 | # and functions return as soon as an error is found. The caller | |
65 | # is also expected to abort on the first error so that debugging | |
66 | # can be done by looking at the end of the output. | |
67 | # | |
68 | # Each function is documented, implemented and tested independently. | |
69 | # When modifying a helper, the test and the documentation are | |
70 | # expected to be updated and it is easier of they are collocated. A | |
71 | # test for a given function can be run with | |
72 | # | |
73 | # ~~~~~~~~~~~~~~~~{.sh} | |
74 | # ceph-helpers.sh TESTS test_get_osds | |
75 | # ~~~~~~~~~~~~~~~~ | |
76 | # | |
77 | # and all the tests (i.e. all functions matching test_*) are run | |
78 | # with: | |
79 | # | |
80 | # ~~~~~~~~~~~~~~~~{.sh} | |
81 | # ceph-helpers.sh TESTS | |
82 | # ~~~~~~~~~~~~~~~~ | |
83 | # | |
84 | # A test function takes a single argument : the directory dedicated | |
85 | # to the tests. It is expected to not create any file outside of this | |
86 | # directory and remove it entirely when it completes successfully. | |
87 | # | |
88 | ||
89 | ||
90 | ## | |
91 | # Cleanup any leftovers found in **dir** via **teardown** | |
92 | # and reset **dir** as an empty environment. | |
93 | # | |
94 | # @param dir path name of the environment | |
95 | # @return 0 on success, 1 on error | |
96 | # | |
97 | function setup() { | |
98 | local dir=$1 | |
99 | teardown $dir || return 1 | |
100 | mkdir -p $dir | |
101 | } | |
102 | ||
103 | function test_setup() { | |
104 | local dir=$dir | |
105 | setup $dir || return 1 | |
106 | test -d $dir || return 1 | |
107 | setup $dir || return 1 | |
108 | test -d $dir || return 1 | |
109 | teardown $dir | |
110 | } | |
111 | ||
112 | ####################################################################### | |
113 | ||
114 | ## | |
115 | # Kill all daemons for which a .pid file exists in **dir** and remove | |
116 | # **dir**. If the file system in which **dir** is btrfs, delete all | |
117 | # subvolumes that relate to it. | |
118 | # | |
119 | # @param dir path name of the environment | |
120 | # @return 0 on success, 1 on error | |
121 | # | |
122 | function teardown() { | |
123 | local dir=$1 | |
124 | kill_daemons $dir KILL | |
125 | if [ `uname` != FreeBSD ] \ | |
126 | && [ $(stat -f -c '%T' .) == "btrfs" ]; then | |
127 | __teardown_btrfs $dir | |
128 | fi | |
129 | rm -fr $dir | |
130 | } | |
131 | ||
132 | function __teardown_btrfs() { | |
133 | local btrfs_base_dir=$1 | |
134 | local btrfs_root=$(df -P . | tail -1 | awk '{print $NF}') | |
135 | local btrfs_dirs=$(cd $btrfs_base_dir; sudo btrfs subvolume list . -t | awk '/^[0-9]/ {print $4}' | grep "$btrfs_base_dir/$btrfs_dir") | |
136 | for subvolume in $btrfs_dirs; do | |
137 | sudo btrfs subvolume delete $btrfs_root/$subvolume | |
138 | done | |
139 | } | |
140 | ||
141 | function test_teardown() { | |
142 | local dir=$dir | |
143 | setup $dir || return 1 | |
144 | teardown $dir || return 1 | |
145 | ! test -d $dir || return 1 | |
146 | } | |
147 | ||
148 | ####################################################################### | |
149 | ||
150 | ## | |
151 | # Sends a signal to a single daemon. | |
152 | # This is a helper function for kill_daemons | |
153 | # | |
154 | # After the daemon is sent **signal**, its actual termination | |
155 | # will be verified by sending it signal 0. If the daemon is | |
156 | # still alive, kill_daemon will pause for a few seconds and | |
157 | # try again. This will repeat for a fixed number of times | |
158 | # before kill_daemon returns on failure. The list of | |
159 | # sleep intervals can be specified as **delays** and defaults | |
160 | # to: | |
161 | # | |
162 | # 0.1 0.2 1 1 1 2 3 5 5 5 10 10 20 60 60 60 120 | |
163 | # | |
164 | # This sequence is designed to run first a very short sleep time (0.1) | |
165 | # if the machine is fast enough and the daemon terminates in a fraction of a | |
166 | # second. The increasing sleep numbers should give plenty of time for | |
167 | # the daemon to die even on the slowest running machine. If a daemon | |
168 | # takes more than a few minutes to stop (the sum of all sleep times), | |
169 | # there probably is no point in waiting more and a number of things | |
170 | # are likely to go wrong anyway: better give up and return on error. | |
171 | # | |
172 | # @param pid the process id to send a signal | |
173 | # @param send_signal the signal to send | |
174 | # @param delays sequence of sleep times before failure | |
175 | # | |
176 | function kill_daemon() { | |
177 | set -x | |
178 | local pid=$(cat $1) | |
179 | local send_signal=$2 | |
180 | local delays=${3:-0.1 0.2 1 1 1 2 3 5 5 5 10 10 20 60 60 60 120} | |
181 | local exit_code=1 | |
182 | for try in $delays ; do | |
183 | if kill -$send_signal $pid 2> /dev/null ; then | |
184 | exit_code=1 | |
185 | else | |
186 | exit_code=0 | |
187 | break | |
188 | fi | |
189 | send_signal=0 | |
190 | sleep $try | |
191 | done; | |
192 | return $exit_code | |
193 | } | |
194 | ||
195 | function test_kill_daemon() { | |
196 | local dir=$1 | |
197 | setup $dir || return 1 | |
198 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
199 | run_mgr $dir x || return 1 | |
200 | run_osd $dir 0 || return 1 | |
201 | ||
202 | name_prefix=osd | |
203 | for pidfile in $(find $dir 2>/dev/null | grep $name_prefix'[^/]*\.pid') ; do | |
204 | # | |
205 | # sending signal 0 won't kill the daemon | |
206 | # waiting just for one second instead of the default schedule | |
207 | # allows us to quickly verify what happens when kill fails | |
208 | # to stop the daemon (i.e. it must return false) | |
209 | # | |
210 | ! kill_daemon $pidfile 0 1 || return 1 | |
211 | # | |
212 | # killing just the osd and verify the mon still is responsive | |
213 | # | |
214 | kill_daemon $pidfile TERM || return 1 | |
215 | done | |
216 | ||
217 | ceph osd dump | grep "osd.0 down" || return 1 | |
218 | ||
219 | name_prefix=mgr | |
220 | for pidfile in $(find $dir 2>/dev/null | grep $name_prefix'[^/]*\.pid') ; do | |
221 | # | |
222 | # kill the mgr | |
223 | # | |
224 | kill_daemon $pidfile TERM || return 1 | |
225 | done | |
226 | ||
227 | name_prefix=mon | |
228 | for pidfile in $(find $dir 2>/dev/null | grep $name_prefix'[^/]*\.pid') ; do | |
229 | # | |
230 | # kill the mon and verify it cannot be reached | |
231 | # | |
232 | kill_daemon $pidfile TERM || return 1 | |
233 | ! timeout 60 ceph --connect-timeout 60 status || return 1 | |
234 | done | |
235 | ||
236 | teardown $dir || return 1 | |
237 | } | |
238 | ||
239 | ## | |
240 | # Kill all daemons for which a .pid file exists in **dir**. Each | |
241 | # daemon is sent a **signal** and kill_daemons waits for it to exit | |
242 | # during a few minutes. By default all daemons are killed. If a | |
243 | # **name_prefix** is provided, only the daemons for which a pid | |
244 | # file is found matching the prefix are killed. See run_osd and | |
245 | # run_mon for more information about the name conventions for | |
246 | # the pid files. | |
247 | # | |
248 | # Send TERM to all daemons : kill_daemons $dir | |
249 | # Send KILL to all daemons : kill_daemons $dir KILL | |
250 | # Send KILL to all osds : kill_daemons $dir KILL osd | |
251 | # Send KILL to osd 1 : kill_daemons $dir KILL osd.1 | |
252 | # | |
253 | # If a daemon is sent the TERM signal and does not terminate | |
254 | # within a few minutes, it will still be running even after | |
255 | # kill_daemons returns. | |
256 | # | |
257 | # If all daemons are kill successfully the function returns 0 | |
258 | # if at least one daemon remains, this is treated as an | |
259 | # error and the function return 1. | |
260 | # | |
261 | # @param dir path name of the environment | |
262 | # @param signal name of the first signal (defaults to TERM) | |
263 | # @param name_prefix only kill match daemons (defaults to all) | |
264 | # @param delays sequence of sleep times before failure | |
265 | # @return 0 on success, 1 on error | |
266 | # | |
267 | function kill_daemons() { | |
268 | local trace=$(shopt -q -o xtrace && echo true || echo false) | |
269 | $trace && shopt -u -o xtrace | |
270 | local dir=$1 | |
271 | local signal=${2:-TERM} | |
272 | local name_prefix=$3 # optional, osd, mon, osd.1 | |
273 | local delays=$4 #optional timing | |
274 | local status=0 | |
275 | local pids="" | |
276 | ||
277 | for pidfile in $(find $dir 2>/dev/null | grep $name_prefix'[^/]*\.pid') ; do | |
278 | run_in_background pids kill_daemon $pidfile $signal $delays | |
279 | done | |
280 | ||
281 | wait_background pids | |
282 | status=$? | |
283 | ||
284 | $trace && shopt -s -o xtrace | |
285 | return $status | |
286 | } | |
287 | ||
288 | function test_kill_daemons() { | |
289 | local dir=$1 | |
290 | setup $dir || return 1 | |
291 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
292 | run_mgr $dir x || return 1 | |
293 | run_osd $dir 0 || return 1 | |
294 | # | |
295 | # sending signal 0 won't kill the daemon | |
296 | # waiting just for one second instead of the default schedule | |
297 | # allows us to quickly verify what happens when kill fails | |
298 | # to stop the daemon (i.e. it must return false) | |
299 | # | |
300 | ! kill_daemons $dir 0 osd 1 || return 1 | |
301 | # | |
302 | # killing just the osd and verify the mon still is responsive | |
303 | # | |
304 | kill_daemons $dir TERM osd || return 1 | |
305 | ceph osd dump | grep "osd.0 down" || return 1 | |
306 | # | |
307 | # kill the mgr | |
308 | # | |
309 | kill_daemons $dir TERM mgr || return 1 | |
310 | # | |
311 | # kill the mon and verify it cannot be reached | |
312 | # | |
313 | kill_daemons $dir TERM || return 1 | |
314 | ! timeout 60 ceph --connect-timeout 60 status || return 1 | |
315 | teardown $dir || return 1 | |
316 | } | |
317 | ||
318 | ####################################################################### | |
319 | ||
320 | ## | |
321 | # Run a monitor by the name mon.**id** with data in **dir**/**id**. | |
322 | # The logs can be found in **dir**/mon.**id**.log and the pid file | |
323 | # is **dir**/mon.**id**.pid and the admin socket is | |
324 | # **dir**/**id**/ceph-mon.**id**.asok. | |
325 | # | |
326 | # The remaining arguments are passed verbatim to ceph-mon --mkfs | |
327 | # and the ceph-mon daemon. | |
328 | # | |
329 | # Two mandatory arguments must be provided: --fsid and --mon-host | |
330 | # Instead of adding them to every call to run_mon, they can be | |
331 | # set in the CEPH_ARGS environment variable to be read implicitly | |
332 | # by every ceph command. | |
333 | # | |
334 | # The CEPH_CONF variable is expected to be set to /dev/null to | |
335 | # only rely on arguments for configuration. | |
336 | # | |
337 | # Examples: | |
338 | # | |
339 | # CEPH_ARGS="--fsid=$(uuidgen) " | |
340 | # CEPH_ARGS+="--mon-host=127.0.0.1:7018 " | |
341 | # run_mon $dir a # spawn a mon and bind port 7018 | |
342 | # run_mon $dir a --debug-filestore=20 # spawn with filestore debugging | |
343 | # | |
344 | # If mon_initial_members is not set, the default rbd pool is deleted | |
345 | # and replaced with a replicated pool with less placement groups to | |
346 | # speed up initialization. If mon_initial_members is set, no attempt | |
347 | # is made to recreate the rbd pool because it would hang forever, | |
348 | # waiting for other mons to join. | |
349 | # | |
350 | # A **dir**/ceph.conf file is created but not meant to be used by any | |
351 | # function. It is convenient for debugging a failure with: | |
352 | # | |
353 | # ceph --conf **dir**/ceph.conf -s | |
354 | # | |
355 | # @param dir path name of the environment | |
356 | # @param id mon identifier | |
357 | # @param ... can be any option valid for ceph-mon | |
358 | # @return 0 on success, 1 on error | |
359 | # | |
360 | function run_mon() { | |
361 | local dir=$1 | |
362 | shift | |
363 | local id=$1 | |
364 | shift | |
365 | local data=$dir/$id | |
366 | ||
367 | ceph-mon \ | |
368 | --id $id \ | |
369 | --mkfs \ | |
370 | --mon-data=$data \ | |
371 | --run-dir=$dir \ | |
372 | "$@" || return 1 | |
373 | ||
374 | ceph-mon \ | |
375 | --id $id \ | |
376 | --mon-osd-full-ratio=.99 \ | |
377 | --mon-data-avail-crit=1 \ | |
378 | --paxos-propose-interval=0.1 \ | |
379 | --osd-crush-chooseleaf-type=0 \ | |
380 | --erasure-code-dir=$CEPH_LIB \ | |
381 | --plugin-dir=$CEPH_LIB \ | |
382 | --debug-mon 20 \ | |
383 | --debug-ms 20 \ | |
384 | --debug-paxos 20 \ | |
385 | --chdir= \ | |
386 | --mon-data=$data \ | |
387 | --log-file=$dir/\$name.log \ | |
388 | --admin-socket=$dir/\$cluster-\$name.asok \ | |
389 | --mon-cluster-log-file=$dir/log \ | |
390 | --run-dir=$dir \ | |
391 | --pid-file=$dir/\$name.pid \ | |
392 | --mon-allow-pool-delete \ | |
393 | "$@" || return 1 | |
394 | ||
395 | cat > $dir/ceph.conf <<EOF | |
396 | [global] | |
397 | fsid = $(get_config mon $id fsid) | |
398 | mon host = $(get_config mon $id mon_host) | |
399 | EOF | |
400 | if test -z "$(get_config mon $id mon_initial_members)" ; then | |
401 | ceph osd pool delete rbd rbd --yes-i-really-really-mean-it || return 1 | |
402 | ceph osd pool create rbd $PG_NUM || return 1 | |
403 | ceph osd set-backfillfull-ratio .99 | |
404 | fi | |
405 | } | |
406 | ||
407 | function test_run_mon() { | |
408 | local dir=$1 | |
409 | ||
410 | setup $dir || return 1 | |
411 | ||
412 | run_mon $dir a --mon-initial-members=a || return 1 | |
413 | # rbd has not been deleted / created, hence it has pool id 0 | |
414 | ceph osd dump | grep "pool 0 'rbd'" || return 1 | |
415 | kill_daemons $dir || return 1 | |
416 | ||
417 | run_mon $dir a || return 1 | |
418 | # rbd has been deleted / created, hence it does not have pool id 0 | |
419 | ! ceph osd dump | grep "pool 0 'rbd'" || return 1 | |
420 | local size=$(CEPH_ARGS='' ceph --format=json daemon $dir/ceph-mon.a.asok \ | |
421 | config get osd_pool_default_size) | |
422 | test "$size" = '{"osd_pool_default_size":"3"}' || return 1 | |
423 | ||
424 | ! CEPH_ARGS='' ceph status || return 1 | |
425 | CEPH_ARGS='' ceph --conf $dir/ceph.conf status || return 1 | |
426 | ||
427 | kill_daemons $dir || return 1 | |
428 | ||
429 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
430 | local size=$(CEPH_ARGS='' ceph --format=json daemon $dir/ceph-mon.a.asok \ | |
431 | config get osd_pool_default_size) | |
432 | test "$size" = '{"osd_pool_default_size":"1"}' || return 1 | |
433 | kill_daemons $dir || return 1 | |
434 | ||
435 | CEPH_ARGS="$CEPH_ARGS --osd_pool_default_size=2" \ | |
436 | run_mon $dir a || return 1 | |
437 | local size=$(CEPH_ARGS='' ceph --format=json daemon $dir/ceph-mon.a.asok \ | |
438 | config get osd_pool_default_size) | |
439 | test "$size" = '{"osd_pool_default_size":"2"}' || return 1 | |
440 | kill_daemons $dir || return 1 | |
441 | ||
442 | teardown $dir || return 1 | |
443 | } | |
444 | ||
445 | ####################################################################### | |
446 | ||
447 | function run_mgr() { | |
448 | local dir=$1 | |
449 | shift | |
450 | local id=$1 | |
451 | shift | |
452 | local data=$dir/$id | |
453 | ||
454 | ceph-mgr \ | |
455 | --id $id \ | |
456 | --erasure-code-dir=$CEPH_LIB \ | |
457 | --plugin-dir=$CEPH_LIB \ | |
458 | --debug-mgr 20 \ | |
459 | --debug-objecter 20 \ | |
460 | --debug-ms 20 \ | |
461 | --debug-paxos 20 \ | |
462 | --chdir= \ | |
463 | --mgr-data=$data \ | |
464 | --log-file=$dir/\$name.log \ | |
465 | --admin-socket=$dir/\$cluster-\$name.asok \ | |
466 | --run-dir=$dir \ | |
467 | --pid-file=$dir/\$name.pid \ | |
468 | "$@" || return 1 | |
469 | } | |
470 | ||
471 | ####################################################################### | |
472 | ||
473 | ## | |
474 | # Create (prepare) and run (activate) an osd by the name osd.**id** | |
475 | # with data in **dir**/**id**. The logs can be found in | |
476 | # **dir**/osd.**id**.log, the pid file is **dir**/osd.**id**.pid and | |
477 | # the admin socket is **dir**/**id**/ceph-osd.**id**.asok. | |
478 | # | |
479 | # The remaining arguments are passed verbatim to ceph-osd. | |
480 | # | |
481 | # Two mandatory arguments must be provided: --fsid and --mon-host | |
482 | # Instead of adding them to every call to run_osd, they can be | |
483 | # set in the CEPH_ARGS environment variable to be read implicitly | |
484 | # by every ceph command. | |
485 | # | |
486 | # The CEPH_CONF variable is expected to be set to /dev/null to | |
487 | # only rely on arguments for configuration. | |
488 | # | |
489 | # The run_osd function creates the OSD data directory with ceph-disk | |
490 | # prepare on the **dir**/**id** directory and relies on the | |
491 | # activate_osd function to run the daemon. | |
492 | # | |
493 | # Examples: | |
494 | # | |
495 | # CEPH_ARGS="--fsid=$(uuidgen) " | |
496 | # CEPH_ARGS+="--mon-host=127.0.0.1:7018 " | |
497 | # run_osd $dir 0 # prepare and activate an osd using the monitor listening on 7018 | |
498 | # | |
499 | # @param dir path name of the environment | |
500 | # @param id osd identifier | |
501 | # @param ... can be any option valid for ceph-osd | |
502 | # @return 0 on success, 1 on error | |
503 | # | |
504 | function run_osd() { | |
505 | local dir=$1 | |
506 | shift | |
507 | local id=$1 | |
508 | shift | |
509 | local osd_data=$dir/$id | |
510 | ||
511 | local ceph_disk_args | |
512 | ceph_disk_args+=" --statedir=$dir" | |
513 | ceph_disk_args+=" --sysconfdir=$dir" | |
514 | ceph_disk_args+=" --prepend-to-path=" | |
515 | ||
516 | mkdir -p $osd_data | |
517 | ceph-disk $ceph_disk_args \ | |
518 | prepare $osd_data || return 1 | |
519 | ||
520 | activate_osd $dir $id "$@" | |
521 | } | |
522 | ||
523 | function run_osd_bluestore() { | |
524 | local dir=$1 | |
525 | shift | |
526 | local id=$1 | |
527 | shift | |
528 | local osd_data=$dir/$id | |
529 | ||
530 | local ceph_disk_args | |
531 | ceph_disk_args+=" --statedir=$dir" | |
532 | ceph_disk_args+=" --sysconfdir=$dir" | |
533 | ceph_disk_args+=" --prepend-to-path=" | |
534 | ||
535 | mkdir -p $osd_data | |
536 | ceph-disk $ceph_disk_args \ | |
537 | prepare --bluestore $osd_data || return 1 | |
538 | ||
539 | local ceph_osd_args | |
540 | ceph_osd_args+=" --enable-experimental-unrecoverable-data-corrupting-features=bluestore" | |
541 | activate_osd $dir $id $ceph_osd_args "$@" | |
542 | } | |
543 | ||
544 | function test_run_osd() { | |
545 | local dir=$1 | |
546 | ||
547 | setup $dir || return 1 | |
548 | ||
549 | run_mon $dir a || return 1 | |
550 | run_mgr $dir x || return 1 | |
551 | ||
552 | run_osd $dir 0 || return 1 | |
553 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $dir//ceph-osd.0.asok \ | |
554 | config get osd_max_backfills) | |
555 | echo "$backfills" | grep --quiet 'osd_max_backfills' || return 1 | |
556 | ||
557 | run_osd $dir 1 --osd-max-backfills 20 || return 1 | |
558 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $dir//ceph-osd.1.asok \ | |
559 | config get osd_max_backfills) | |
560 | test "$backfills" = '{"osd_max_backfills":"20"}' || return 1 | |
561 | ||
562 | CEPH_ARGS="$CEPH_ARGS --osd-max-backfills 30" run_osd $dir 2 || return 1 | |
563 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $dir//ceph-osd.2.asok \ | |
564 | config get osd_max_backfills) | |
565 | test "$backfills" = '{"osd_max_backfills":"30"}' || return 1 | |
566 | ||
567 | teardown $dir || return 1 | |
568 | } | |
569 | ||
570 | ####################################################################### | |
571 | ||
572 | ## | |
573 | # Shutdown and remove all traces of the osd by the name osd.**id**. | |
574 | # | |
575 | # The OSD is shutdown with the TERM signal. It is then removed from | |
576 | # the auth list, crush map, osd map etc and the files associated with | |
577 | # it are also removed. | |
578 | # | |
579 | # @param dir path name of the environment | |
580 | # @param id osd identifier | |
581 | # @return 0 on success, 1 on error | |
582 | # | |
583 | function destroy_osd() { | |
584 | local dir=$1 | |
585 | local id=$2 | |
586 | ||
587 | kill_daemons $dir TERM osd.$id || return 1 | |
588 | ceph osd out osd.$id || return 1 | |
589 | ceph auth del osd.$id || return 1 | |
590 | ceph osd crush remove osd.$id || return 1 | |
591 | ceph osd rm $id || return 1 | |
592 | teardown $dir/$id || return 1 | |
593 | rm -fr $dir/$id | |
594 | } | |
595 | ||
596 | function test_destroy_osd() { | |
597 | local dir=$1 | |
598 | ||
599 | setup $dir || return 1 | |
600 | run_mon $dir a || return 1 | |
601 | run_mgr $dir x || return 1 | |
602 | run_osd $dir 0 || return 1 | |
603 | destroy_osd $dir 0 || return 1 | |
604 | ! ceph osd dump | grep "osd.$id " || return 1 | |
605 | teardown $dir || return 1 | |
606 | } | |
607 | ||
608 | ####################################################################### | |
609 | ||
610 | ## | |
611 | # Run (activate) an osd by the name osd.**id** with data in | |
612 | # **dir**/**id**. The logs can be found in **dir**/osd.**id**.log, | |
613 | # the pid file is **dir**/osd.**id**.pid and the admin socket is | |
614 | # **dir**/**id**/ceph-osd.**id**.asok. | |
615 | # | |
616 | # The remaining arguments are passed verbatim to ceph-osd. | |
617 | # | |
618 | # Two mandatory arguments must be provided: --fsid and --mon-host | |
619 | # Instead of adding them to every call to activate_osd, they can be | |
620 | # set in the CEPH_ARGS environment variable to be read implicitly | |
621 | # by every ceph command. | |
622 | # | |
623 | # The CEPH_CONF variable is expected to be set to /dev/null to | |
624 | # only rely on arguments for configuration. | |
625 | # | |
626 | # The activate_osd function expects a valid OSD data directory | |
627 | # in **dir**/**id**, either just created via run_osd or re-using | |
628 | # one left by a previous run of ceph-osd. The ceph-osd daemon is | |
629 | # run indirectly via ceph-disk activate. | |
630 | # | |
631 | # The activate_osd function blocks until the monitor reports the osd | |
632 | # up. If it fails to do so within $TIMEOUT seconds, activate_osd | |
633 | # fails. | |
634 | # | |
635 | # Examples: | |
636 | # | |
637 | # CEPH_ARGS="--fsid=$(uuidgen) " | |
638 | # CEPH_ARGS+="--mon-host=127.0.0.1:7018 " | |
639 | # activate_osd $dir 0 # activate an osd using the monitor listening on 7018 | |
640 | # | |
641 | # @param dir path name of the environment | |
642 | # @param id osd identifier | |
643 | # @param ... can be any option valid for ceph-osd | |
644 | # @return 0 on success, 1 on error | |
645 | # | |
646 | function activate_osd() { | |
647 | local dir=$1 | |
648 | shift | |
649 | local id=$1 | |
650 | shift | |
651 | local osd_data=$dir/$id | |
652 | ||
653 | local ceph_disk_args | |
654 | ceph_disk_args+=" --statedir=$dir" | |
655 | ceph_disk_args+=" --sysconfdir=$dir" | |
656 | ceph_disk_args+=" --prepend-to-path=" | |
657 | ||
658 | local ceph_args="$CEPH_ARGS" | |
659 | ceph_args+=" --enable-experimental-unrecoverable-data-corrupting-features=bluestore" | |
660 | ceph_args+=" --osd-failsafe-full-ratio=.99" | |
661 | ceph_args+=" --osd-journal-size=100" | |
662 | ceph_args+=" --osd-scrub-load-threshold=2000" | |
663 | ceph_args+=" --osd-data=$osd_data" | |
664 | ceph_args+=" --chdir=" | |
665 | ceph_args+=" --erasure-code-dir=$CEPH_LIB" | |
666 | ceph_args+=" --plugin-dir=$CEPH_LIB" | |
667 | ceph_args+=" --osd-class-dir=$CEPH_LIB" | |
668 | ceph_args+=" --run-dir=$dir" | |
669 | ceph_args+=" --debug-osd=20" | |
670 | ceph_args+=" --log-file=$dir/\$name.log" | |
671 | ceph_args+=" --pid-file=$dir/\$name.pid" | |
672 | ceph_args+=" --osd-max-object-name-len 460" | |
673 | ceph_args+=" --osd-max-object-namespace-len 64" | |
674 | ceph_args+=" " | |
675 | ceph_args+="$@" | |
676 | mkdir -p $osd_data | |
677 | CEPH_ARGS="$ceph_args " ceph-disk $ceph_disk_args \ | |
678 | activate \ | |
679 | --mark-init=none \ | |
680 | $osd_data || return 1 | |
681 | ||
682 | [ "$id" = "$(cat $osd_data/whoami)" ] || return 1 | |
683 | ||
684 | wait_for_osd up $id || return 1 | |
685 | } | |
686 | ||
687 | function test_activate_osd() { | |
688 | local dir=$1 | |
689 | ||
690 | setup $dir || return 1 | |
691 | ||
692 | run_mon $dir a || return 1 | |
693 | run_mgr $dir x || return 1 | |
694 | ||
695 | run_osd $dir 0 || return 1 | |
696 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $dir//ceph-osd.0.asok \ | |
697 | config get osd_max_backfills) | |
698 | echo "$backfills" | grep --quiet 'osd_max_backfills' || return 1 | |
699 | ||
700 | kill_daemons $dir TERM osd || return 1 | |
701 | ||
702 | activate_osd $dir 0 --osd-max-backfills 20 || return 1 | |
703 | local backfills=$(CEPH_ARGS='' ceph --format=json daemon $dir//ceph-osd.0.asok \ | |
704 | config get osd_max_backfills) | |
705 | test "$backfills" = '{"osd_max_backfills":"20"}' || return 1 | |
706 | ||
707 | teardown $dir || return 1 | |
708 | } | |
709 | ||
710 | ####################################################################### | |
711 | ||
712 | ## | |
713 | # Wait until the OSD **id** is either up or down, as specified by | |
714 | # **state**. It fails after $TIMEOUT seconds. | |
715 | # | |
716 | # @param state either up or down | |
717 | # @param id osd identifier | |
718 | # @return 0 on success, 1 on error | |
719 | # | |
720 | function wait_for_osd() { | |
721 | local state=$1 | |
722 | local id=$2 | |
723 | ||
724 | status=1 | |
725 | for ((i=0; i < $TIMEOUT; i++)); do | |
726 | echo $i | |
727 | if ! ceph osd dump | grep "osd.$id $state"; then | |
728 | sleep 1 | |
729 | else | |
730 | status=0 | |
731 | break | |
732 | fi | |
733 | done | |
734 | return $status | |
735 | } | |
736 | ||
737 | function test_wait_for_osd() { | |
738 | local dir=$1 | |
739 | setup $dir || return 1 | |
740 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
741 | run_mgr $dir x || return 1 | |
742 | run_osd $dir 0 || return 1 | |
743 | wait_for_osd up 0 || return 1 | |
744 | kill_daemons $dir TERM osd || return 1 | |
745 | wait_for_osd down 0 || return 1 | |
746 | ( TIMEOUT=1 ; ! wait_for_osd up 0 ) || return 1 | |
747 | teardown $dir || return 1 | |
748 | } | |
749 | ||
750 | ####################################################################### | |
751 | ||
752 | ## | |
753 | # Display the list of OSD ids supporting the **objectname** stored in | |
754 | # **poolname**, as reported by ceph osd map. | |
755 | # | |
756 | # @param poolname an existing pool | |
757 | # @param objectname an objectname (may or may not exist) | |
758 | # @param STDOUT white space separated list of OSD ids | |
759 | # @return 0 on success, 1 on error | |
760 | # | |
761 | function get_osds() { | |
762 | local poolname=$1 | |
763 | local objectname=$2 | |
764 | ||
765 | local osds=$(ceph --format xml osd map $poolname $objectname 2>/dev/null | \ | |
766 | $XMLSTARLET sel -t -m "//acting/osd" -v . -o ' ') | |
767 | # get rid of the trailing space | |
768 | echo $osds | |
769 | } | |
770 | ||
771 | function test_get_osds() { | |
772 | local dir=$1 | |
773 | ||
774 | setup $dir || return 1 | |
775 | run_mon $dir a --osd_pool_default_size=2 || return 1 | |
776 | run_mgr $dir x || return 1 | |
777 | run_osd $dir 0 || return 1 | |
778 | run_osd $dir 1 || return 1 | |
779 | wait_for_clean || return 1 | |
780 | get_osds rbd GROUP | grep --quiet '^[0-1] [0-1]$' || return 1 | |
781 | teardown $dir || return 1 | |
782 | } | |
783 | ||
784 | ####################################################################### | |
785 | ||
786 | ## | |
787 | # Wait for the monitor to form quorum (optionally, of size N) | |
788 | # | |
789 | # @param timeout duration (lower-bound) to wait for quorum to be formed | |
790 | # @param quorumsize size of quorum to wait for | |
791 | # @return 0 on success, 1 on error | |
792 | # | |
793 | function wait_for_quorum() { | |
794 | local timeout=$1 | |
795 | local quorumsize=$2 | |
796 | ||
797 | if [[ -z "$timeout" ]]; then | |
798 | timeout=300 | |
799 | fi | |
800 | ||
801 | if [[ -z "$quorumsize" ]]; then | |
802 | timeout $timeout ceph mon_status --format=json >&/dev/null || return 1 | |
803 | return 0 | |
804 | fi | |
805 | ||
806 | no_quorum=1 | |
807 | wait_until=$((`date +%s` + $timeout)) | |
808 | while [[ $(date +%s) -lt $wait_until ]]; do | |
809 | jqfilter='.quorum | length == '$quorumsize | |
810 | jqinput="$(timeout $timeout ceph mon_status --format=json 2>/dev/null)" | |
811 | res=$(echo $jqinput | jq "$jqfilter") | |
812 | if [[ "$res" == "true" ]]; then | |
813 | no_quorum=0 | |
814 | break | |
815 | fi | |
816 | done | |
817 | return $no_quorum | |
818 | } | |
819 | ||
820 | ####################################################################### | |
821 | ||
822 | ## | |
823 | # Return the PG of supporting the **objectname** stored in | |
824 | # **poolname**, as reported by ceph osd map. | |
825 | # | |
826 | # @param poolname an existing pool | |
827 | # @param objectname an objectname (may or may not exist) | |
828 | # @param STDOUT a PG | |
829 | # @return 0 on success, 1 on error | |
830 | # | |
831 | function get_pg() { | |
832 | local poolname=$1 | |
833 | local objectname=$2 | |
834 | ||
835 | ceph --format xml osd map $poolname $objectname 2>/dev/null | \ | |
836 | $XMLSTARLET sel -t -m "//pgid" -v . -n | |
837 | } | |
838 | ||
839 | function test_get_pg() { | |
840 | local dir=$1 | |
841 | ||
842 | setup $dir || return 1 | |
843 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
844 | run_mgr $dir x || return 1 | |
845 | run_osd $dir 0 || return 1 | |
846 | wait_for_clean || return 1 | |
847 | get_pg rbd GROUP | grep --quiet '^[0-9]\.[0-9a-f][0-9a-f]*$' || return 1 | |
848 | teardown $dir || return 1 | |
849 | } | |
850 | ||
851 | ####################################################################### | |
852 | ||
853 | ## | |
854 | # Return the value of the **config**, obtained via the config get command | |
855 | # of the admin socket of **daemon**.**id**. | |
856 | # | |
857 | # @param daemon mon or osd | |
858 | # @param id mon or osd ID | |
859 | # @param config the configuration variable name as found in config_opts.h | |
860 | # @param STDOUT the config value | |
861 | # @return 0 on success, 1 on error | |
862 | # | |
863 | function get_config() { | |
864 | local daemon=$1 | |
865 | local id=$2 | |
866 | local config=$3 | |
867 | ||
868 | CEPH_ARGS='' \ | |
869 | ceph --format xml daemon $dir/ceph-$daemon.$id.asok \ | |
870 | config get $config 2> /dev/null | \ | |
871 | $XMLSTARLET sel -t -m "//$config" -v . -n | |
872 | } | |
873 | ||
874 | function test_get_config() { | |
875 | local dir=$1 | |
876 | ||
877 | # override the default config using command line arg and check it | |
878 | setup $dir || return 1 | |
879 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
880 | test $(get_config mon a osd_pool_default_size) = 1 || return 1 | |
881 | run_mgr $dir x || return 1 | |
882 | run_osd $dir 0 --osd_max_scrubs=3 || return 1 | |
883 | test $(get_config osd 0 osd_max_scrubs) = 3 || return 1 | |
884 | teardown $dir || return 1 | |
885 | } | |
886 | ||
887 | ####################################################################### | |
888 | ||
889 | ## | |
890 | # Set the **config** to specified **value**, via the config set command | |
891 | # of the admin socket of **daemon**.**id** | |
892 | # | |
893 | # @param daemon mon or osd | |
894 | # @param id mon or osd ID | |
895 | # @param config the configuration variable name as found in config_opts.h | |
896 | # @param value the config value | |
897 | # @return 0 on success, 1 on error | |
898 | # | |
899 | function set_config() { | |
900 | local daemon=$1 | |
901 | local id=$2 | |
902 | local config=$3 | |
903 | local value=$4 | |
904 | ||
905 | CEPH_ARGS='' \ | |
906 | ceph --format xml daemon $dir/ceph-$daemon.$id.asok \ | |
907 | config set $config $value 2> /dev/null | \ | |
908 | $XMLSTARLET sel -Q -t -m "//success" -v . | |
909 | } | |
910 | ||
911 | function test_set_config() { | |
912 | local dir=$1 | |
913 | ||
914 | setup $dir || return 1 | |
915 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
916 | test $(get_config mon a ms_crc_header) = true || return 1 | |
917 | set_config mon a ms_crc_header false || return 1 | |
918 | test $(get_config mon a ms_crc_header) = false || return 1 | |
919 | set_config mon a ms_crc_header true || return 1 | |
920 | test $(get_config mon a ms_crc_header) = true || return 1 | |
921 | teardown $dir || return 1 | |
922 | } | |
923 | ||
924 | ####################################################################### | |
925 | ||
926 | ## | |
927 | # Return the OSD id of the primary OSD supporting the **objectname** | |
928 | # stored in **poolname**, as reported by ceph osd map. | |
929 | # | |
930 | # @param poolname an existing pool | |
931 | # @param objectname an objectname (may or may not exist) | |
932 | # @param STDOUT the primary OSD id | |
933 | # @return 0 on success, 1 on error | |
934 | # | |
935 | function get_primary() { | |
936 | local poolname=$1 | |
937 | local objectname=$2 | |
938 | ||
939 | ceph --format xml osd map $poolname $objectname 2>/dev/null | \ | |
940 | $XMLSTARLET sel -t -m "//acting_primary" -v . -n | |
941 | } | |
942 | ||
943 | function test_get_primary() { | |
944 | local dir=$1 | |
945 | ||
946 | setup $dir || return 1 | |
947 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
948 | local osd=0 | |
949 | run_mgr $dir x || return 1 | |
950 | run_osd $dir $osd || return 1 | |
951 | wait_for_clean || return 1 | |
952 | test $(get_primary rbd GROUP) = $osd || return 1 | |
953 | teardown $dir || return 1 | |
954 | } | |
955 | ||
956 | ####################################################################### | |
957 | ||
958 | ## | |
959 | # Return the id of any OSD supporting the **objectname** stored in | |
960 | # **poolname**, as reported by ceph osd map, except the primary. | |
961 | # | |
962 | # @param poolname an existing pool | |
963 | # @param objectname an objectname (may or may not exist) | |
964 | # @param STDOUT the OSD id | |
965 | # @return 0 on success, 1 on error | |
966 | # | |
967 | function get_not_primary() { | |
968 | local poolname=$1 | |
969 | local objectname=$2 | |
970 | ||
971 | local primary=$(get_primary $poolname $objectname) | |
972 | ceph --format xml osd map $poolname $objectname 2>/dev/null | \ | |
973 | $XMLSTARLET sel -t -m "//acting/osd[not(.='$primary')]" -v . -n | \ | |
974 | head -1 | |
975 | } | |
976 | ||
977 | function test_get_not_primary() { | |
978 | local dir=$1 | |
979 | ||
980 | setup $dir || return 1 | |
981 | run_mon $dir a --osd_pool_default_size=2 || return 1 | |
982 | run_mgr $dir x || return 1 | |
983 | run_osd $dir 0 || return 1 | |
984 | run_osd $dir 1 || return 1 | |
985 | wait_for_clean || return 1 | |
986 | local primary=$(get_primary rbd GROUP) | |
987 | local not_primary=$(get_not_primary rbd GROUP) | |
988 | test $not_primary != $primary || return 1 | |
989 | test $not_primary = 0 -o $not_primary = 1 || return 1 | |
990 | teardown $dir || return 1 | |
991 | } | |
992 | ||
993 | ####################################################################### | |
994 | ||
995 | ## | |
996 | # Run ceph-objectstore-tool against the OSD **id** using the data path | |
997 | # **dir**. The OSD is killed with TERM prior to running | |
998 | # ceph-objectstore-tool because access to the data path is | |
999 | # exclusive. The OSD is restarted after the command completes. The | |
1000 | # objectstore_tool returns after all PG are active+clean again. | |
1001 | # | |
1002 | # @param dir the data path of the OSD | |
1003 | # @param id the OSD id | |
1004 | # @param ... arguments to ceph-objectstore-tool | |
1005 | # @param STDIN the input of ceph-objectstore-tool | |
1006 | # @param STDOUT the output of ceph-objectstore-tool | |
1007 | # @return 0 on success, 1 on error | |
1008 | # | |
1009 | # The value of $ceph_osd_args will be passed to restarted osds | |
1010 | # | |
1011 | function objectstore_tool() { | |
1012 | local dir=$1 | |
1013 | shift | |
1014 | local id=$1 | |
1015 | shift | |
1016 | local osd_data=$dir/$id | |
1017 | ||
1018 | local osd_type=$(cat $osd_data/type) | |
1019 | ||
1020 | kill_daemons $dir TERM osd.$id >&2 < /dev/null || return 1 | |
1021 | ||
1022 | local journal_args | |
1023 | if [ "$objectstore_type" == "filestore" ]; then | |
1024 | journal_args=" --journal-path $osd_data/journal" | |
1025 | fi | |
1026 | ceph-objectstore-tool \ | |
1027 | --enable-experimental-unrecoverable-data-corrupting-features=bluestore \ | |
1028 | --data-path $osd_data \ | |
1029 | $journal_args \ | |
1030 | "$@" || return 1 | |
1031 | activate_osd $dir $id $ceph_osd_args >&2 || return 1 | |
1032 | wait_for_clean >&2 | |
1033 | } | |
1034 | ||
1035 | function test_objectstore_tool() { | |
1036 | local dir=$1 | |
1037 | ||
1038 | setup $dir || return 1 | |
1039 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1040 | local osd=0 | |
1041 | run_mgr $dir x || return 1 | |
1042 | run_osd $dir $osd || return 1 | |
1043 | wait_for_clean || return 1 | |
1044 | rados --pool rbd put GROUP /etc/group || return 1 | |
1045 | objectstore_tool $dir $osd GROUP get-bytes | \ | |
1046 | diff - /etc/group | |
1047 | ! objectstore_tool $dir $osd NOTEXISTS get-bytes || return 1 | |
1048 | teardown $dir || return 1 | |
1049 | } | |
1050 | ||
1051 | ####################################################################### | |
1052 | ||
1053 | ## | |
1054 | # Predicate checking if there is an ongoing recovery in the | |
1055 | # cluster. If any of the recovering_{keys,bytes,objects}_per_sec | |
1056 | # counters are reported by ceph status, it means recovery is in | |
1057 | # progress. | |
1058 | # | |
1059 | # @return 0 if recovery in progress, 1 otherwise | |
1060 | # | |
1061 | function get_is_making_recovery_progress() { | |
1062 | local progress=$(ceph --format xml status 2>/dev/null | \ | |
1063 | $XMLSTARLET sel \ | |
1064 | -t -m "//pgmap/recovering_keys_per_sec" -v . -o ' ' \ | |
1065 | -t -m "//pgmap/recovering_bytes_per_sec" -v . -o ' ' \ | |
1066 | -t -m "//pgmap/recovering_objects_per_sec" -v .) | |
1067 | test -n "$progress" | |
1068 | } | |
1069 | ||
1070 | function test_get_is_making_recovery_progress() { | |
1071 | local dir=$1 | |
1072 | ||
1073 | setup $dir || return 1 | |
1074 | run_mon $dir a || return 1 | |
1075 | run_mgr $dir x || return 1 | |
1076 | ! get_is_making_recovery_progress || return 1 | |
1077 | teardown $dir || return 1 | |
1078 | } | |
1079 | ||
1080 | ####################################################################### | |
1081 | ||
1082 | ## | |
1083 | # Return the number of active PGs in the cluster. A PG is active if | |
1084 | # ceph pg dump pgs reports it both **active** and **clean** and that | |
1085 | # not **stale**. | |
1086 | # | |
1087 | # @param STDOUT the number of active PGs | |
1088 | # @return 0 on success, 1 on error | |
1089 | # | |
1090 | function get_num_active_clean() { | |
1091 | local expression="(" | |
1092 | expression+="contains(.,'active') and " | |
1093 | expression+="contains(.,'clean') and " | |
1094 | expression+="not(contains(.,'stale'))" | |
1095 | expression+=")" | |
1096 | # xmlstarlet 1.3.0 (which is on Ubuntu precise) | |
1097 | # add extra new lines that must be ignored with | |
1098 | # grep -v '^$' | |
1099 | ceph --format xml pg dump pgs 2>/dev/null | \ | |
1100 | $XMLSTARLET sel -t -m "//pg_stat/state[$expression]" -v . -n | \ | |
1101 | grep -cv '^$' | |
1102 | } | |
1103 | ||
1104 | function test_get_num_active_clean() { | |
1105 | local dir=$1 | |
1106 | ||
1107 | setup $dir || return 1 | |
1108 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1109 | run_mgr $dir x || return 1 | |
1110 | run_osd $dir 0 || return 1 | |
1111 | wait_for_clean || return 1 | |
1112 | local num_active_clean=$(get_num_active_clean) | |
1113 | test "$num_active_clean" = $PG_NUM || return 1 | |
1114 | teardown $dir || return 1 | |
1115 | } | |
1116 | ||
1117 | ####################################################################### | |
1118 | ||
1119 | ## | |
1120 | # Return the number of PGs in the cluster, according to | |
1121 | # ceph pg dump pgs. | |
1122 | # | |
1123 | # @param STDOUT the number of PGs | |
1124 | # @return 0 on success, 1 on error | |
1125 | # | |
1126 | function get_num_pgs() { | |
1127 | ceph --format xml status 2>/dev/null | \ | |
1128 | $XMLSTARLET sel -t -m "//pgmap/num_pgs" -v . | |
1129 | } | |
1130 | ||
1131 | function test_get_num_pgs() { | |
1132 | local dir=$1 | |
1133 | ||
1134 | setup $dir || return 1 | |
1135 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1136 | run_mgr $dir x || return 1 | |
1137 | run_osd $dir 0 || return 1 | |
1138 | wait_for_clean || return 1 | |
1139 | local num_pgs=$(get_num_pgs) | |
1140 | test "$num_pgs" -gt 0 || return 1 | |
1141 | teardown $dir || return 1 | |
1142 | } | |
1143 | ||
1144 | ####################################################################### | |
1145 | ||
1146 | ## | |
1147 | # Return the date and time of the last completed scrub for **pgid**, | |
1148 | # as reported by ceph pg dump pgs. Note that a repair also sets this | |
1149 | # date. | |
1150 | # | |
1151 | # @param pgid the id of the PG | |
1152 | # @param STDOUT the date and time of the last scrub | |
1153 | # @return 0 on success, 1 on error | |
1154 | # | |
1155 | function get_last_scrub_stamp() { | |
1156 | local pgid=$1 | |
1157 | local sname=${2:-last_scrub_stamp} | |
1158 | ceph --format xml pg dump pgs 2>/dev/null | \ | |
1159 | $XMLSTARLET sel -t -m "//pg_stat[pgid='$pgid']/$sname" -v . | |
1160 | } | |
1161 | ||
1162 | function test_get_last_scrub_stamp() { | |
1163 | local dir=$1 | |
1164 | ||
1165 | setup $dir || return 1 | |
1166 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1167 | run_mgr $dir x || return 1 | |
1168 | run_osd $dir 0 || return 1 | |
1169 | wait_for_clean || return 1 | |
1170 | stamp=$(get_last_scrub_stamp 1.0) | |
1171 | test -n "$stamp" || return 1 | |
1172 | teardown $dir || return 1 | |
1173 | } | |
1174 | ||
1175 | ####################################################################### | |
1176 | ||
1177 | ## | |
1178 | # Predicate checking if the cluster is clean, i.e. all of its PGs are | |
1179 | # in a clean state (see get_num_active_clean for a definition). | |
1180 | # | |
1181 | # @return 0 if the cluster is clean, 1 otherwise | |
1182 | # | |
1183 | function is_clean() { | |
1184 | num_pgs=$(get_num_pgs) | |
1185 | test $num_pgs != 0 || return 1 | |
1186 | test $(get_num_active_clean) = $num_pgs || return 1 | |
1187 | } | |
1188 | ||
1189 | function test_is_clean() { | |
1190 | local dir=$1 | |
1191 | ||
1192 | setup $dir || return 1 | |
1193 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1194 | run_mgr $dir x || return 1 | |
1195 | run_osd $dir 0 || return 1 | |
1196 | wait_for_clean || return 1 | |
1197 | is_clean || return 1 | |
1198 | teardown $dir || return 1 | |
1199 | } | |
1200 | ||
1201 | ####################################################################### | |
1202 | ||
1203 | ## | |
1204 | # Return a list of numbers that are increasingly larger and whose | |
1205 | # total is **timeout** seconds. It can be used to have short sleep | |
1206 | # delay while waiting for an event on a fast machine. But if running | |
1207 | # very slowly the larger delays avoid stressing the machine even | |
1208 | # further or spamming the logs. | |
1209 | # | |
1210 | # @param timeout sum of all delays, in seconds | |
1211 | # @return a list of sleep delays | |
1212 | # | |
1213 | function get_timeout_delays() { | |
1214 | local trace=$(shopt -q -o xtrace && echo true || echo false) | |
1215 | $trace && shopt -u -o xtrace | |
1216 | local timeout=$1 | |
1217 | local first_step=${2:-1} | |
1218 | ||
1219 | local i | |
1220 | local total="0" | |
1221 | i=$first_step | |
1222 | while test "$(echo $total + $i \<= $timeout | bc -l)" = "1"; do | |
1223 | echo -n "$i " | |
1224 | total=$(echo $total + $i | bc -l) | |
1225 | i=$(echo $i \* 2 | bc -l) | |
1226 | done | |
1227 | if test "$(echo $total \< $timeout | bc -l)" = "1"; then | |
1228 | echo -n $(echo $timeout - $total | bc -l) | |
1229 | fi | |
1230 | $trace && shopt -s -o xtrace | |
1231 | } | |
1232 | ||
1233 | function test_get_timeout_delays() { | |
1234 | test "$(get_timeout_delays 1)" = "1 " || return 1 | |
1235 | test "$(get_timeout_delays 5)" = "1 2 2" || return 1 | |
1236 | test "$(get_timeout_delays 6)" = "1 2 3" || return 1 | |
1237 | test "$(get_timeout_delays 7)" = "1 2 4 " || return 1 | |
1238 | test "$(get_timeout_delays 8)" = "1 2 4 1" || return 1 | |
1239 | test "$(get_timeout_delays 1 .1)" = ".1 .2 .4 .3" || return 1 | |
1240 | test "$(get_timeout_delays 1.5 .1)" = ".1 .2 .4 .8 " || return 1 | |
1241 | test "$(get_timeout_delays 5 .1)" = ".1 .2 .4 .8 1.6 1.9" || return 1 | |
1242 | test "$(get_timeout_delays 6 .1)" = ".1 .2 .4 .8 1.6 2.9" || return 1 | |
1243 | test "$(get_timeout_delays 6.3 .1)" = ".1 .2 .4 .8 1.6 3.2 " || return 1 | |
1244 | test "$(get_timeout_delays 20 .1)" = ".1 .2 .4 .8 1.6 3.2 6.4 7.3" || return 1 | |
1245 | } | |
1246 | ||
1247 | ####################################################################### | |
1248 | ||
1249 | ## | |
1250 | # Wait until the cluster becomes clean or if it does not make progress | |
1251 | # for $TIMEOUT seconds. | |
1252 | # Progress is measured either via the **get_is_making_recovery_progress** | |
1253 | # predicate or if the number of clean PGs changes (as returned by get_num_active_clean) | |
1254 | # | |
1255 | # @return 0 if the cluster is clean, 1 otherwise | |
1256 | # | |
1257 | function wait_for_clean() { | |
1258 | local num_active_clean=-1 | |
1259 | local cur_active_clean | |
1260 | local -a delays=($(get_timeout_delays $TIMEOUT .1)) | |
1261 | local -i loop=0 | |
1262 | test $(get_num_pgs) != 0 || return 1 | |
1263 | ||
1264 | while true ; do | |
1265 | # Comparing get_num_active_clean & get_num_pgs is used to determine | |
1266 | # if the cluster is clean. That's almost an inline of is_clean() to | |
1267 | # get more performance by avoiding multiple calls of get_num_active_clean. | |
1268 | cur_active_clean=$(get_num_active_clean) | |
1269 | test $cur_active_clean = $(get_num_pgs) && break | |
1270 | if test $cur_active_clean != $num_active_clean ; then | |
1271 | loop=0 | |
1272 | num_active_clean=$cur_active_clean | |
1273 | elif get_is_making_recovery_progress ; then | |
1274 | loop=0 | |
1275 | elif (( $loop >= ${#delays[*]} )) ; then | |
1276 | ceph report | |
1277 | return 1 | |
1278 | fi | |
1279 | sleep ${delays[$loop]} | |
1280 | loop+=1 | |
1281 | done | |
1282 | return 0 | |
1283 | } | |
1284 | ||
1285 | function test_wait_for_clean() { | |
1286 | local dir=$1 | |
1287 | ||
1288 | setup $dir || return 1 | |
1289 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1290 | run_mgr $dir x || return 1 | |
1291 | ! TIMEOUT=1 wait_for_clean || return 1 | |
1292 | run_osd $dir 0 || return 1 | |
1293 | wait_for_clean || return 1 | |
1294 | teardown $dir || return 1 | |
1295 | } | |
1296 | ||
1297 | ####################################################################### | |
1298 | ||
1299 | ## | |
1300 | # Wait until the cluster becomes HEALTH_OK again or if it does not make progress | |
1301 | # for $TIMEOUT seconds. | |
1302 | # | |
1303 | # @return 0 if the cluster is HEALTHY, 1 otherwise | |
1304 | # | |
1305 | function wait_for_health() { | |
1306 | local grepstr=$1 | |
1307 | local -a delays=($(get_timeout_delays $TIMEOUT .1)) | |
1308 | local -i loop=0 | |
1309 | ||
1310 | while ! ceph health detail | grep "$grepstr" ; do | |
1311 | if (( $loop >= ${#delays[*]} )) ; then | |
1312 | ceph health detail | |
1313 | return 1 | |
1314 | fi | |
1315 | sleep ${delays[$loop]} | |
1316 | loop+=1 | |
1317 | done | |
1318 | } | |
1319 | ||
1320 | function wait_for_health_ok() { | |
1321 | wait_for_health "HEALTH_OK" || return 1 | |
1322 | } | |
1323 | ||
1324 | function test_wait_for_health_ok() { | |
1325 | local dir=$1 | |
1326 | ||
1327 | setup $dir || return 1 | |
1328 | run_mon $dir a --osd_pool_default_size=1 --osd_failsafe_full_ratio=.99 --mon_pg_warn_min_per_osd=0 || return 1 | |
1329 | run_mgr $dir x || return 1 | |
1330 | ! TIMEOUT=1 wait_for_health_ok || return 1 | |
1331 | run_osd $dir 0 || return 1 | |
1332 | wait_for_health_ok || return 1 | |
1333 | teardown $dir || return 1 | |
1334 | } | |
1335 | ||
1336 | ||
1337 | ####################################################################### | |
1338 | ||
1339 | ## | |
1340 | # Run repair on **pgid** and wait until it completes. The repair | |
1341 | # function will fail if repair does not complete within $TIMEOUT | |
1342 | # seconds. | |
1343 | # | |
1344 | # @param pgid the id of the PG | |
1345 | # @return 0 on success, 1 on error | |
1346 | # | |
1347 | function repair() { | |
1348 | local pgid=$1 | |
1349 | local last_scrub=$(get_last_scrub_stamp $pgid) | |
1350 | ceph pg repair $pgid | |
1351 | wait_for_scrub $pgid "$last_scrub" | |
1352 | } | |
1353 | ||
1354 | function test_repair() { | |
1355 | local dir=$1 | |
1356 | ||
1357 | setup $dir || return 1 | |
1358 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1359 | run_mgr $dir x || return 1 | |
1360 | run_osd $dir 0 || return 1 | |
1361 | wait_for_clean || return 1 | |
1362 | repair 1.0 || return 1 | |
1363 | kill_daemons $dir KILL osd || return 1 | |
1364 | ! TIMEOUT=1 repair 1.0 || return 1 | |
1365 | teardown $dir || return 1 | |
1366 | } | |
1367 | ####################################################################### | |
1368 | ||
1369 | ## | |
1370 | # Run scrub on **pgid** and wait until it completes. The pg_scrub | |
1371 | # function will fail if repair does not complete within $TIMEOUT | |
1372 | # seconds. The pg_scrub is complete whenever the | |
1373 | # **get_last_scrub_stamp** function reports a timestamp different from | |
1374 | # the one stored before starting the scrub. | |
1375 | # | |
1376 | # @param pgid the id of the PG | |
1377 | # @return 0 on success, 1 on error | |
1378 | # | |
1379 | function pg_scrub() { | |
1380 | local pgid=$1 | |
1381 | local last_scrub=$(get_last_scrub_stamp $pgid) | |
1382 | ceph pg scrub $pgid | |
1383 | wait_for_scrub $pgid "$last_scrub" | |
1384 | } | |
1385 | ||
1386 | function pg_deep_scrub() { | |
1387 | local pgid=$1 | |
1388 | local last_scrub=$(get_last_scrub_stamp $pgid last_deep_scrub_stamp) | |
1389 | ceph pg deep-scrub $pgid | |
1390 | wait_for_scrub $pgid "$last_scrub" last_deep_scrub_stamp | |
1391 | } | |
1392 | ||
1393 | function test_pg_scrub() { | |
1394 | local dir=$1 | |
1395 | ||
1396 | setup $dir || return 1 | |
1397 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1398 | run_mgr $dir x || return 1 | |
1399 | run_osd $dir 0 || return 1 | |
1400 | wait_for_clean || return 1 | |
1401 | pg_scrub 1.0 || return 1 | |
1402 | kill_daemons $dir KILL osd || return 1 | |
1403 | ! TIMEOUT=1 pg_scrub 1.0 || return 1 | |
1404 | teardown $dir || return 1 | |
1405 | } | |
1406 | ||
1407 | ####################################################################### | |
1408 | ||
1409 | ## | |
1410 | # Run the *command* and expect it to fail (i.e. return a non zero status). | |
1411 | # The output (stderr and stdout) is stored in a temporary file in *dir* | |
1412 | # and is expected to contain the string *expected*. | |
1413 | # | |
1414 | # Return 0 if the command failed and the string was found. Otherwise | |
1415 | # return 1 and cat the full output of the command on stderr for debug. | |
1416 | # | |
1417 | # @param dir temporary directory to store the output | |
1418 | # @param expected string to look for in the output | |
1419 | # @param command ... the command and its arguments | |
1420 | # @return 0 on success, 1 on error | |
1421 | # | |
1422 | ||
1423 | function expect_failure() { | |
1424 | local dir=$1 | |
1425 | shift | |
1426 | local expected="$1" | |
1427 | shift | |
1428 | local success | |
1429 | ||
1430 | if "$@" > $dir/out 2>&1 ; then | |
1431 | success=true | |
1432 | else | |
1433 | success=false | |
1434 | fi | |
1435 | ||
1436 | if $success || ! grep --quiet "$expected" $dir/out ; then | |
1437 | cat $dir/out >&2 | |
1438 | return 1 | |
1439 | else | |
1440 | return 0 | |
1441 | fi | |
1442 | } | |
1443 | ||
1444 | function test_expect_failure() { | |
1445 | local dir=$1 | |
1446 | ||
1447 | setup $dir || return 1 | |
1448 | expect_failure $dir FAIL bash -c 'echo FAIL ; exit 1' || return 1 | |
1449 | # the command did not fail | |
1450 | ! expect_failure $dir FAIL bash -c 'echo FAIL ; exit 0' > $dir/out || return 1 | |
1451 | grep --quiet FAIL $dir/out || return 1 | |
1452 | # the command failed but the output does not contain the expected string | |
1453 | ! expect_failure $dir FAIL bash -c 'echo UNEXPECTED ; exit 1' > $dir/out || return 1 | |
1454 | ! grep --quiet FAIL $dir/out || return 1 | |
1455 | teardown $dir || return 1 | |
1456 | } | |
1457 | ||
1458 | ####################################################################### | |
1459 | ||
1460 | ## | |
1461 | # Given the *last_scrub*, wait for scrub to happen on **pgid**. It | |
1462 | # will fail if scrub does not complete within $TIMEOUT seconds. The | |
1463 | # repair is complete whenever the **get_last_scrub_stamp** function | |
1464 | # reports a timestamp different from the one given in argument. | |
1465 | # | |
1466 | # @param pgid the id of the PG | |
1467 | # @param last_scrub timestamp of the last scrub for *pgid* | |
1468 | # @return 0 on success, 1 on error | |
1469 | # | |
1470 | function wait_for_scrub() { | |
1471 | local pgid=$1 | |
1472 | local last_scrub="$2" | |
1473 | local sname=${3:-last_scrub_stamp} | |
1474 | ||
1475 | for ((i=0; i < $TIMEOUT; i++)); do | |
1476 | if test "$last_scrub" != "$(get_last_scrub_stamp $pgid $sname)" ; then | |
1477 | return 0 | |
1478 | fi | |
1479 | sleep 1 | |
1480 | done | |
1481 | return 1 | |
1482 | } | |
1483 | ||
1484 | function test_wait_for_scrub() { | |
1485 | local dir=$1 | |
1486 | ||
1487 | setup $dir || return 1 | |
1488 | run_mon $dir a --osd_pool_default_size=1 || return 1 | |
1489 | run_mgr $dir x || return 1 | |
1490 | run_osd $dir 0 || return 1 | |
1491 | wait_for_clean || return 1 | |
1492 | local pgid=1.0 | |
1493 | ceph pg repair $pgid | |
1494 | local last_scrub=$(get_last_scrub_stamp $pgid) | |
1495 | wait_for_scrub $pgid "$last_scrub" || return 1 | |
1496 | kill_daemons $dir KILL osd || return 1 | |
1497 | last_scrub=$(get_last_scrub_stamp $pgid) | |
1498 | ! TIMEOUT=1 wait_for_scrub $pgid "$last_scrub" || return 1 | |
1499 | teardown $dir || return 1 | |
1500 | } | |
1501 | ||
1502 | ####################################################################### | |
1503 | ||
1504 | ## | |
1505 | # Return 0 if the erasure code *plugin* is available, 1 otherwise. | |
1506 | # | |
1507 | # @param plugin erasure code plugin | |
1508 | # @return 0 on success, 1 on error | |
1509 | # | |
1510 | ||
1511 | function erasure_code_plugin_exists() { | |
1512 | local plugin=$1 | |
1513 | local status | |
1514 | local grepstr | |
1515 | local s | |
1516 | case `uname` in | |
1517 | FreeBSD) grepstr="Cannot open.*$plugin" ;; | |
1518 | *) grepstr="$plugin.*No such file" ;; | |
1519 | esac | |
1520 | ||
1521 | s=$(ceph osd erasure-code-profile set TESTPROFILE plugin=$plugin 2>&1) | |
1522 | local status=$? | |
1523 | if [ $status -eq 0 ]; then | |
1524 | ceph osd erasure-code-profile rm TESTPROFILE | |
1525 | elif ! echo $s | grep --quiet "$grepstr" ; then | |
1526 | status=1 | |
1527 | # display why the string was rejected. | |
1528 | echo $s | |
1529 | fi | |
1530 | return $status | |
1531 | } | |
1532 | ||
1533 | function test_erasure_code_plugin_exists() { | |
1534 | local dir=$1 | |
1535 | ||
1536 | setup $dir || return 1 | |
1537 | run_mon $dir a || return 1 | |
1538 | run_mgr $dir x || return 1 | |
1539 | erasure_code_plugin_exists jerasure || return 1 | |
1540 | ! erasure_code_plugin_exists FAKE || return 1 | |
1541 | teardown $dir || return 1 | |
1542 | } | |
1543 | ||
1544 | ####################################################################### | |
1545 | ||
1546 | ## | |
1547 | # Display all log files from **dir** on stdout. | |
1548 | # | |
1549 | # @param dir directory in which all data is stored | |
1550 | # | |
1551 | ||
1552 | function display_logs() { | |
1553 | local dir=$1 | |
1554 | ||
1555 | find $dir -maxdepth 1 -name '*.log' | \ | |
1556 | while read file ; do | |
1557 | echo "======================= $file" | |
1558 | cat $file | |
1559 | done | |
1560 | } | |
1561 | ||
1562 | function test_display_logs() { | |
1563 | local dir=$1 | |
1564 | ||
1565 | setup $dir || return 1 | |
1566 | run_mon $dir a || return 1 | |
1567 | kill_daemons $dir || return 1 | |
1568 | display_logs $dir > $dir/log.out | |
1569 | grep --quiet mon.a.log $dir/log.out || return 1 | |
1570 | teardown $dir || return 1 | |
1571 | } | |
1572 | ||
1573 | ####################################################################### | |
1574 | ## | |
1575 | # Spawn a command in background and save the pid in the variable name | |
1576 | # passed in argument. To make the output reading easier, the output is | |
1577 | # prepend with the process id. | |
1578 | # | |
1579 | # Example: | |
1580 | # pids1="" | |
1581 | # run_in_background pids1 bash -c 'sleep 1; exit 1' | |
1582 | # | |
1583 | # @param pid_variable the variable name (not value) where the pids will be stored | |
1584 | # @param ... the command to execute | |
1585 | # @return only the pid_variable output should be considered and used with **wait_background** | |
1586 | # | |
1587 | function run_in_background() { | |
1588 | local pid_variable=$1 | |
1589 | shift; | |
1590 | # Execute the command and prepend the output with its pid | |
1591 | # We enforce to return the exit status of the command and not the awk one. | |
1592 | ("$@" |& awk '{ a[i++] = $0 }END{for (i = 0; i in a; ++i) { print "'$$': " a[i]} }'; return ${PIPESTATUS[0]}) >&2 & | |
1593 | eval "$pid_variable+=\" $!\"" | |
1594 | } | |
1595 | ||
1596 | function test_run_in_background() { | |
1597 | local pids | |
1598 | run_in_background pids sleep 1 | |
1599 | run_in_background pids sleep 1 | |
1600 | test $(echo $pids | wc -w) = 2 || return 1 | |
1601 | wait $pids || return 1 | |
1602 | } | |
1603 | ||
1604 | ####################################################################### | |
1605 | ## | |
1606 | # Wait for pids running in background to complete. | |
1607 | # This function is usually used after a **run_in_background** call | |
1608 | # Example: | |
1609 | # pids1="" | |
1610 | # run_in_background pids1 bash -c 'sleep 1; exit 1' | |
1611 | # wait_background pids1 | |
1612 | # | |
1613 | # @param pids The variable name that contains the active PIDS. Set as empty at then end of the function. | |
1614 | # @return returns 1 if at least one process exits in error unless returns 0 | |
1615 | # | |
1616 | function wait_background() { | |
1617 | # We extract the PIDS from the variable name | |
1618 | pids=${!1} | |
1619 | ||
1620 | return_code=0 | |
1621 | for pid in $pids; do | |
1622 | if ! wait $pid; then | |
1623 | # If one process failed then return 1 | |
1624 | return_code=1 | |
1625 | fi | |
1626 | done | |
1627 | ||
1628 | # We empty the variable reporting that all process ended | |
1629 | eval "$1=''" | |
1630 | ||
1631 | return $return_code | |
1632 | } | |
1633 | ||
1634 | ||
1635 | function test_wait_background() { | |
1636 | local pids="" | |
1637 | run_in_background pids bash -c "sleep 1; exit 1" | |
1638 | run_in_background pids bash -c "sleep 2; exit 0" | |
1639 | wait_background pids | |
1640 | if [ $? -ne 1 ]; then return 1; fi | |
1641 | ||
1642 | run_in_background pids bash -c "sleep 1; exit 0" | |
1643 | run_in_background pids bash -c "sleep 2; exit 0" | |
1644 | wait_background pids | |
1645 | if [ $? -ne 0 ]; then return 1; fi | |
1646 | ||
1647 | if [ ! -z "$pids" ]; then return 1; fi | |
1648 | } | |
1649 | ||
1650 | ####################################################################### | |
1651 | ||
1652 | ## | |
1653 | # Call the **run** function (which must be defined by the caller) with | |
1654 | # the **dir** argument followed by the caller argument list. | |
1655 | # | |
1656 | # If the **run** function returns on error, all logs found in **dir** | |
1657 | # are displayed for diagnostic purposes. | |
1658 | # | |
1659 | # **teardown** function is called when the **run** function returns | |
1660 | # (on success or on error), to cleanup leftovers. The CEPH_CONF is set | |
1661 | # to /dev/null and CEPH_ARGS is unset so that the tests are protected from | |
1662 | # external interferences. | |
1663 | # | |
1664 | # It is the responsibility of the **run** function to call the | |
1665 | # **setup** function to prepare the test environment (create a temporary | |
1666 | # directory etc.). | |
1667 | # | |
1668 | # The shell is required (via PS4) to display the function and line | |
1669 | # number whenever a statement is executed to help debugging. | |
1670 | # | |
1671 | # @param dir directory in which all data is stored | |
1672 | # @param ... arguments passed transparently to **run** | |
1673 | # @return 0 on success, 1 on error | |
1674 | # | |
1675 | function main() { | |
1676 | local dir=td/$1 | |
1677 | shift | |
1678 | ||
1679 | shopt -s -o xtrace | |
1680 | PS4='${BASH_SOURCE[0]}:$LINENO: ${FUNCNAME[0]}: ' | |
1681 | ||
1682 | export PATH=${CEPH_BUILD_VIRTUALENV}/ceph-disk-virtualenv/bin:${CEPH_BUILD_VIRTUALENV}/ceph-detect-init-virtualenv/bin:.:$PATH # make sure program from sources are preferred | |
1683 | #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 | |
1684 | ||
1685 | export CEPH_CONF=/dev/null | |
1686 | unset CEPH_ARGS | |
1687 | ||
1688 | local code | |
1689 | if run $dir "$@" ; then | |
1690 | code=0 | |
1691 | else | |
1692 | display_logs $dir | |
1693 | code=1 | |
1694 | fi | |
1695 | teardown $dir || return 1 | |
1696 | return $code | |
1697 | } | |
1698 | ||
1699 | ####################################################################### | |
1700 | ||
1701 | function run_tests() { | |
1702 | shopt -s -o xtrace | |
1703 | PS4='${BASH_SOURCE[0]}:$LINENO: ${FUNCNAME[0]}: ' | |
1704 | ||
1705 | export PATH=${CEPH_BUILD_VIRTUALENV}/ceph-disk-virtualenv/bin:${CEPH_BUILD_VIRTUALENV}/ceph-detect-init-virtualenv/bin:.:$PATH # make sure program from sources are preferred | |
1706 | #export PATH=$CEPH_ROOT/src/ceph-disk/virtualenv/bin:$CEPH_ROOT/src/ceph-detect-init/virtualenv/bin:.:$PATH # make sure program from sources are preferred | |
1707 | ||
1708 | export CEPH_MON="127.0.0.1:7109" # git grep '\<7109\>' : there must be only one | |
1709 | export CEPH_ARGS | |
1710 | CEPH_ARGS+="--fsid=$(uuidgen) --auth-supported=none " | |
1711 | CEPH_ARGS+="--mon-host=$CEPH_MON " | |
1712 | export CEPH_CONF=/dev/null | |
1713 | ||
1714 | local funcs=${@:-$(set | sed -n -e 's/^\(test_[0-9a-z_]*\) .*/\1/p')} | |
1715 | local dir=td/ceph-helpers | |
1716 | ||
1717 | for func in $funcs ; do | |
1718 | $func $dir || return 1 | |
1719 | done | |
1720 | } | |
1721 | ||
1722 | if test "$1" = TESTS ; then | |
1723 | shift | |
1724 | run_tests "$@" | |
1725 | fi | |
1726 | ||
1727 | # Local Variables: | |
1728 | # compile-command: "cd ../../src ; make -j4 && ../qa/workunits/ceph-helpers.sh TESTS # test_get_config" | |
1729 | # End: |