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