xref: /illumos-gate/usr/src/test/zfs-tests/tests/functional/pool_checkpoint/pool_checkpoint.kshlib (revision c65ebfc7045424bd04a6c7719a27b0ad3399ad54)
1#
2# This file and its contents are supplied under the terms of the
3# Common Development and Distribution License ("CDDL"), version 1.0.
4# You may only use this file in accordance with the terms of version
5# 1.0 of the CDDL.
6#
7# A full copy of the text of the CDDL should have accompanied this
8# source.  A copy of the CDDL is also available via the Internet at
9# http://www.illumos.org/license/CDDL.
10#
11
12#
13# Copyright (c) 2017, 2018 by Delphix. All rights reserved.
14#
15
16. $STF_SUITE/include/libtest.shlib
17. $STF_SUITE/tests/functional/removal/removal.kshlib
18
19#
20# In general all the tests related to the pool checkpoint can
21# be divided into two categories. TESTS that verify features
22# provided by the checkpoint (e.g. checkpoint_rewind) and tests
23# that stress-test the checkpoint (e.g. checkpoint_big_rewind).
24#
25# For the first group we don't really care about the size of
26# the pool or the individual file sizes within the filesystems.
27# This is why these tests run directly on pools that use a
28# "real disk vdev" (meaning not a file based one). These tests
29# use the $TESTPOOL pool that is created on top of $TESTDISK.
30# This pool is refered to as the "test pool" and thus all
31# the tests of this group use the testpool-related functions of
32# this file (not the nested_pools ones).
33#
34# For the second group we generally try to bring the pool to its
35# limits by increasing fragmentation, filling all allocatable
36# space, attempting to use vdevs that the checkpoint spacemap
37# cannot represent, etc. For these tests we need to control
38# almost all parameters of the pool and the vdevs that back it
39# so we create them based on file-based vdevs that we carefully
40# create within the $TESTPOOL pool. So most of these tests, in
41# order to create this nested pool sctructure, generally start
42# like this:
43# 1] We create the test pool ($TESTPOOL).
44# 2] We create a filesystem and we populate it with files of
45#    some predetermined size.
46# 3] We use those files as vdevs for the pool that the test
47#    will use ($NESTEDPOOL).
48# 4] Go on and let the test run and operate on $NESTEDPOOL.
49#
50
51#
52# These disks are used to back $TESTPOOL
53#
54TESTDISK="$(echo $DISKS | cut -d' ' -f1)"
55EXTRATESTDISK="$(echo $DISKS | cut -d' ' -f2)"
56
57FS0=$TESTPOOL/$TESTFS
58FS1=$TESTPOOL/$TESTFS1
59FS2=$TESTPOOL/$TESTFS2
60
61FS0FILE=/$FS0/$TESTFILE0
62FS1FILE=/$FS1/$TESTFILE1
63FS2FILE=/$FS2/$TESTFILE2
64
65#
66# The following are created within $TESTPOOL and
67# will be used to back $NESTEDPOOL
68#
69DISKFS=$TESTPOOL/disks
70FILEDISKDIR=/$DISKFS
71FILEDISK1=/$DISKFS/dsk1
72FILEDISK2=/$DISKFS/dsk2
73FILEDISKS="$FILEDISK1 $FILEDISK2"
74
75#
76# $NESTEDPOOL related variables
77#
78NESTEDPOOL=nestedpool
79NESTEDFS0=$NESTEDPOOL/$TESTFS
80NESTEDFS1=$NESTEDPOOL/$TESTFS1
81NESTEDFS2=$NESTEDPOOL/$TESTFS2
82NESTEDFS0FILE=/$NESTEDFS0/$TESTFILE0
83NESTEDFS1FILE=/$NESTEDFS1/$TESTFILE1
84NESTEDFS2FILE=/$NESTEDFS2/$TESTFILE2
85
86#
87# In the tests that stress-test the pool (second category
88# mentioned above), there exist some that need to bring
89# fragmentation at high percentages in a relatively short
90# period of time. In order to do that we set the following
91# parameters:
92#
93# * We use two disks of 1G each, to create a pool of size 2G.
94#   The point is that 2G is not small nor large, and we also
95#   want to have 2 disks to introduce indirect vdevs on our
96#   setup.
97# * We enable compression and set the record size of all
98#   filesystems to 8K. The point of compression is to
99#   ensure that we are not filling up the whole pool (that's
100#   what checkpoint_capacity is for), and the specific
101#   record size is set to match the block size of randwritecomp
102#   which is used to increase fragmentation by writing on
103#   files.
104# * We always have 2 big files present of 512M each, which
105#   should account for 40%~50% capacity by the end of each
106#   test with fragmentation around 50~60%.
107# * At each file we attempt to do enough random writes to
108#   touch every offset twice on average.
109#
110# Note that the amount of random writes per files are based
111# on the following calculation:
112#
113# ((512M / 8K) * 3) * 2 = ~400000
114#
115# Given that the file is 512M and one write is 8K, we would
116# need (512M / 8K) writes to go through the whole file.
117# Assuming though that each write has a compression ratio of
118# 3, then we want 3 times that to cover the same amount of
119# space. Finally, we multiply that by 2 since our goal is to
120# touch each offset twice on average.
121#
122# Examples of those tests are checkpoint_big_rewind and
123# checkpoint_discard_busy.
124#
125FILEDISKSIZE=1g
126DISKSIZE=1g
127BIGFILESIZE=512M
128RANDOMWRITES=400000
129
130
131#
132# Assumes create_test_pool has been called beforehand.
133#
134function setup_nested_pool
135{
136	log_must zfs create $DISKFS
137
138	log_must mkfile $DISKSIZE $FILEDISK1
139	log_must mkfile $DISKSIZE $FILEDISK2
140
141	log_must zpool create $NESTEDPOOL $FILEDISKS
142}
143
144function setup_test_pool
145{
146	log_must zpool create $TESTPOOL "$TESTDISK"
147}
148
149function setup_nested_pools
150{
151	setup_test_pool
152	setup_nested_pool
153}
154
155function cleanup_nested_pool
156{
157	log_must zpool destroy $NESTEDPOOL
158	log_must rm -f $FILEDISKS
159}
160
161function cleanup_test_pool
162{
163	log_must zpool destroy $TESTPOOL
164}
165
166function cleanup_nested_pools
167{
168	cleanup_nested_pool
169	cleanup_test_pool
170}
171
172#
173# Remove and re-add each vdev to ensure that data is
174# moved between disks and indirect mappings are created
175#
176function introduce_indirection
177{
178	for disk in ${FILEDISKS[@]}; do
179		log_must zpool remove $NESTEDPOOL $disk
180		log_must wait_for_removal $NESTEDPOOL
181		log_mustnot vdevs_in_pool $NESTEDPOOL $disk
182		log_must zpool add $NESTEDPOOL $disk
183	done
184}
185
186FILECONTENTS0="Can't wait to be checkpointed!"
187FILECONTENTS1="Can't wait to be checkpointed too!"
188NEWFILECONTENTS0="I survived after the checkpoint!"
189NEWFILECONTENTS2="I was born after the checkpoint!"
190
191function populate_test_pool
192{
193	log_must zfs create -o compression=lz4 -o recordsize=8k $FS0
194	log_must zfs create -o compression=lz4 -o recordsize=8k $FS1
195
196	echo $FILECONTENTS0 > $FS0FILE
197	echo $FILECONTENTS1 > $FS1FILE
198}
199
200function populate_nested_pool
201{
202	log_must zfs create -o compression=lz4 -o recordsize=8k $NESTEDFS0
203	log_must zfs create -o compression=lz4 -o recordsize=8k $NESTEDFS1
204
205	echo $FILECONTENTS0 > $NESTEDFS0FILE
206	echo $FILECONTENTS1 > $NESTEDFS1FILE
207}
208
209function test_verify_pre_checkpoint_state
210{
211	log_must zfs list $FS0
212	log_must zfs list $FS1
213	log_must [ "$(cat $FS0FILE)" = "$FILECONTENTS0" ]
214	log_must [ "$(cat $FS1FILE)" = "$FILECONTENTS1" ]
215
216	#
217	# If we've opened the checkpointed state of the
218	# pool as read-only without rewinding on-disk we
219	# can't really use zdb on it.
220	#
221	if [[ "$1" != "ro-check" ]] ; then
222		log_must zdb $TESTPOOL
223	fi
224
225	#
226	# Ensure post-checkpoint state is not present
227	#
228	log_mustnot zfs list $FS2
229	log_mustnot [ "$(cat $FS0FILE)" = "$NEWFILECONTENTS0" ]
230}
231
232function nested_verify_pre_checkpoint_state
233{
234	log_must zfs list $NESTEDFS0
235	log_must zfs list $NESTEDFS1
236	log_must [ "$(cat $NESTEDFS0FILE)" = "$FILECONTENTS0" ]
237	log_must [ "$(cat $NESTEDFS1FILE)" = "$FILECONTENTS1" ]
238
239	#
240	# If we've opened the checkpointed state of the
241	# pool as read-only without rewinding on-disk we
242	# can't really use zdb on it.
243	#
244	if [[ "$1" != "ro-check" ]] ; then
245		log_must zdb $NESTEDPOOL
246	fi
247
248	#
249	# Ensure post-checkpoint state is not present
250	#
251	log_mustnot zfs list $NESTEDFS2
252	log_mustnot [ "$(cat $NESTEDFS0FILE)" = "$NEWFILECONTENTS0" ]
253}
254
255function test_change_state_after_checkpoint
256{
257	log_must zfs destroy $FS1
258	log_must zfs create -o compression=lz4 -o recordsize=8k $FS2
259
260	echo $NEWFILECONTENTS0 > $FS0FILE
261	echo $NEWFILECONTENTS2 > $FS2FILE
262}
263
264function nested_change_state_after_checkpoint
265{
266	log_must zfs destroy $NESTEDFS1
267	log_must zfs create -o compression=lz4 -o recordsize=8k $NESTEDFS2
268
269	echo $NEWFILECONTENTS0 > $NESTEDFS0FILE
270	echo $NEWFILECONTENTS2 > $NESTEDFS2FILE
271}
272
273function test_verify_post_checkpoint_state
274{
275	log_must zfs list $FS0
276	log_must zfs list $FS2
277	log_must [ "$(cat $FS0FILE)" = "$NEWFILECONTENTS0" ]
278	log_must [ "$(cat $FS2FILE)" = "$NEWFILECONTENTS2" ]
279
280	log_must zdb $TESTPOOL
281
282	#
283	# Ensure pre-checkpointed state that was removed post-checkpoint
284	# is not present
285	#
286	log_mustnot zfs list $FS1
287	log_mustnot [ "$(cat $FS0FILE)" = "$FILECONTENTS0" ]
288}
289
290function fragment_before_checkpoint
291{
292	populate_nested_pool
293	log_must mkfile -n $BIGFILESIZE $NESTEDFS0FILE
294	log_must mkfile -n $BIGFILESIZE $NESTEDFS1FILE
295	log_must randwritecomp $NESTEDFS0FILE $RANDOMWRITES
296	log_must randwritecomp $NESTEDFS1FILE $RANDOMWRITES
297
298	#
299	# Display fragmentation on test log
300	#
301	log_must zpool list -v
302}
303
304function fragment_after_checkpoint_and_verify
305{
306	log_must zfs destroy $NESTEDFS1
307	log_must zfs create -o compression=lz4 -o recordsize=8k $NESTEDFS2
308	log_must mkfile -n $BIGFILESIZE $NESTEDFS2FILE
309	log_must randwritecomp $NESTEDFS0FILE $RANDOMWRITES
310	log_must randwritecomp $NESTEDFS2FILE $RANDOMWRITES
311
312	#
313	# Display fragmentation on test log
314	#
315	log_must zpool list -v
316
317	log_must zdb $NESTEDPOOL
318	log_must zdb -kc $NESTEDPOOL
319}
320
321function wait_discard_finish
322{
323	typeset pool="$1"
324
325	typeset status
326	status=$(zpool status $pool | grep "checkpoint:")
327	while [ "" != "$status" ]; do
328		sleep 5
329		status=$(zpool status $pool | grep "checkpoint:")
330	done
331}
332
333function test_wait_discard_finish
334{
335	wait_discard_finish $TESTPOOL
336}
337
338function nested_wait_discard_finish
339{
340	wait_discard_finish $NESTEDPOOL
341}
342
343#
344# Creating the setup for the second group of tests mentioned in
345# block comment of this file can take some time as we are doing
346# random writes to raise capacity and fragmentation before taking
347# the checkpoint. Thus we create this setup once and save the
348# disks of the nested pool in a temporary directory where we can
349# reuse it for each test that requires that setup.
350#
351SAVEDPOOLDIR="/var/tmp/ckpoint_saved_pool"
352
353function test_group_premake_nested_pools
354{
355	setup_nested_pools
356
357	#
358	# Populate and fragment the pool.
359	#
360	fragment_before_checkpoint
361
362	#
363	# Export and save the pool for other tests.
364	#
365	log_must zpool export $NESTEDPOOL
366	log_must mkdir $SAVEDPOOLDIR
367	log_must cp $FILEDISKS $SAVEDPOOLDIR
368
369	#
370	# Reimport pool to be destroyed by
371	# cleanup_nested_pools function
372	#
373	log_must zpool import -d $FILEDISKDIR $NESTEDPOOL
374}
375
376function test_group_destroy_saved_pool
377{
378	log_must rm -rf $SAVEDPOOLDIR
379}
380
381#
382# Recreate nested pool setup from saved pool.
383#
384function setup_nested_pool_state
385{
386	setup_test_pool
387
388	log_must zfs create $DISKFS
389	log_must cp $SAVEDPOOLDIR/* $FILEDISKDIR
390
391	log_must zpool import -d $FILEDISKDIR $NESTEDPOOL
392}
393