1#!/usr/bin/ksh -p
2
3#
4# This file and its contents are supplied under the terms of the
5# Common Development and Distribution License ("CDDL"), version 1.0.
6# You may only use this file in accordance with the terms of version
7# 1.0 of the CDDL.
8#
9# A full copy of the text of the CDDL should have accompanied this
10# source.  A copy of the CDDL is also available via the Internet at
11# http://www.illumos.org/license/CDDL.
12#
13
14#
15# Copyright (c) 2016 by Delphix. All rights reserved.
16#
17
18. $STF_SUITE/tests/functional/cli_root/zpool_import/zpool_import.kshlib
19
20#
21# DESCRIPTION:
22#	It should be possible to rewind a pool beyond a device replacement.
23#
24# STRATEGY:
25#	1. Create a pool.
26#	2. Generate files and remember their md5sum.
27#	3. Sync a few times and note last synced txg.
28#	4. Take a snapshot to make sure old blocks are not overwritten.
29#	5. Initiate device replacement and export the pool. Special care must
30#	   be taken so that resilvering doesn't complete before the export.
31#	6. Test 1: Rewind pool to noted txg and then verify data checksums.
32#	   Import it read-only so that we do not overwrite blocks in later txgs.
33#	7. Re-import pool at latest txg and let the replacement finish.
34#	8. Export the pool an remove the new device - we shouldn't need it.
35#	9. Test 2: Rewind pool to noted txg and then verify data checksums.
36#
37# STRATEGY TO SLOW DOWN RESILVERING:
38#	1. Reduce zfs_txg_timeout, which controls how long can we resilver for
39#	   each sync.
40#	2. Add data to pool
41#	3. Re-import the pool so that data isn't cached
42#	4. Use zinject to slow down device I/O
43#	5. Trigger the resilvering
44#	6. Use spa freeze to stop writing to the pool.
45#	7. Clear zinject events (needed to export the pool)
46#	8. Export the pool
47#
48# DISCLAIMER:
49#	This test can fail since nothing guarantees that old MOS blocks aren't
50#	overwritten. Snapshots protect datasets and data files but not the MOS.
51#	sync_some_data_a_few_times interleaves file data and MOS data for a few
52#	txgs, thus increasing the odds that some txgs will have their MOS data
53#	left untouched.
54#
55
56verify_runnable "global"
57
58ZFS_TXG_TIMEOUT=""
59
60function custom_cleanup
61{
62	# Revert zfs_txg_timeout to defaults
63	[[ -n ZFS_TXG_TIMEOUT ]] &&
64	    log_must set_zfs_txg_timeout $ZFS_TXG_TIMEOUT
65	log_must rm -rf $BACKUP_DEVICE_DIR
66	log_must set_tunable32 zfs_scan_suspend_progress 0
67	cleanup
68}
69
70log_onexit custom_cleanup
71
72function test_replace_vdev
73{
74	typeset poolcreate="$1"
75	typeset replacevdev="$2"
76	typeset replaceby="$3"
77	typeset poolfinalstate="$4"
78	typeset zinjectdevices="$5"
79
80	log_note "$0: pool '$poolcreate', replace $replacevdev by $replaceby."
81
82	log_must zpool create $TESTPOOL1 $poolcreate
83
84	# generate data and checksum it
85	log_must generate_data $TESTPOOL1 $MD5FILE
86
87	# add more data so that resilver takes longer
88	log_must write_some_data $TESTPOOL1
89
90	# Syncing a few times while writing new data increases the odds that
91	# MOS metadata for some of the txgs will survive.
92	log_must sync_some_data_a_few_times $TESTPOOL1
93	typeset txg
94	txg=$(get_last_txg_synced $TESTPOOL1)
95	log_must zfs snapshot -r $TESTPOOL1@snap1
96
97	# This should not free original data.
98	log_must overwrite_data $TESTPOOL1 ""
99
100	log_must zpool export $TESTPOOL1
101	log_must zpool import -d $DEVICE_DIR $TESTPOOL1
102
103	# Ensure resilvering doesn't complete.
104	log_must set_tunable32 zfs_scan_suspend_progress 1
105	log_must zpool replace $TESTPOOL1 $replacevdev $replaceby
106
107	# Confirm pool is still replacing
108	log_must pool_is_replacing $TESTPOOL1
109	log_must zpool export $TESTPOOL1
110	log_must set_tunable32 zfs_scan_suspend_progress 0
111
112	############################################################
113	# Test 1: rewind while device is resilvering.
114	# Import read only to avoid overwriting more recent blocks.
115	############################################################
116	log_must zpool import -d $DEVICE_DIR -o readonly=on -T $txg $TESTPOOL1
117	log_must check_pool_config $TESTPOOL1 "$poolcreate"
118
119	log_must verify_data_md5sums $MD5FILE
120
121	log_must zpool export $TESTPOOL1
122
123	# Import pool at latest txg to finish the resilvering
124	log_must zpool import -d $DEVICE_DIR $TESTPOOL1
125	log_must overwrite_data $TESTPOOL1 ""
126	log_must wait_for_pool_config $TESTPOOL1 "$poolfinalstate"
127	log_must zpool export $TESTPOOL1
128
129	# Move out the new device
130	log_must mv $replaceby $BACKUP_DEVICE_DIR/
131
132	############################################################
133	# Test 2: rewind after device has been replaced.
134	# Import read-write since we won't need the pool anymore.
135	############################################################
136	log_must zpool import -d $DEVICE_DIR -T $txg $TESTPOOL1
137	log_must check_pool_config $TESTPOOL1 "$poolcreate"
138
139	log_must verify_data_md5sums $MD5FILE
140
141	# Cleanup
142	log_must zpool destroy $TESTPOOL1
143	# Restore the device we moved out
144	log_must mv "$BACKUP_DEVICE_DIR/$(basename $replaceby)" $DEVICE_DIR/
145	# Fast way to clear vdev labels
146	log_must zpool create -f $TESTPOOL2 $VDEV0 $VDEV1 $VDEV2 $VDEV3 $VDEV4
147	log_must zpool destroy $TESTPOOL2
148
149	log_note ""
150}
151
152log_must mkdir $BACKUP_DEVICE_DIR
153# Make the devices bigger to reduce chances of overwriting MOS metadata.
154increase_device_sizes $(( FILE_SIZE * 4 ))
155
156# We set zfs_txg_timeout to 1 to reduce resilvering time at each sync.
157ZFS_TXG_TIMEOUT=$(get_zfs_txg_timeout)
158set_zfs_txg_timeout 1
159
160test_replace_vdev "$VDEV0 $VDEV1" \
161    "$VDEV1" "$VDEV2" \
162    "$VDEV0 $VDEV2" \
163    "$VDEV0 $VDEV1"
164
165test_replace_vdev "mirror $VDEV0 $VDEV1" \
166	"$VDEV1" "$VDEV2" \
167	"mirror $VDEV0 $VDEV2" \
168	"$VDEV0 $VDEV1"
169
170test_replace_vdev "raidz $VDEV0 $VDEV1 $VDEV2" \
171	"$VDEV1" "$VDEV3" \
172	"raidz $VDEV0 $VDEV3 $VDEV2" \
173	"$VDEV0 $VDEV1 $VDEV2"
174
175set_zfs_txg_timeout $ZFS_TXG_TIMEOUT
176
177log_pass "zpool import rewind after device replacement passed."
178