9075 Improve ZFS pool import/load process and corrupted pool recovery
[unleashed.git] / usr / src / test / zfs-tests / tests / functional / cli_root / zpool_import / import_cachefile_device_replaced.ksh
blob74957770cb9d1b452bdb169cdc48a89b2f26301a
1 #!/usr/bin/ksh -p
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.
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.
15 # Copyright (c) 2016 by Delphix. All rights reserved.
18 . $STF_SUITE/tests/functional/cli_root/zpool_import/zpool_import.kshlib
21 # DESCRIPTION:
22 # A pool should be importable using an outdated cachefile that is unaware
23 # of a zpool replace operation at different stages in time.
25 # STRATEGY:
26 # 1. Create a pool with some devices and an alternate cachefile.
27 # 2. Backup the cachefile.
28 # 3. Initiate device replacement, backup cachefile again and export pool.
29 # Special care must be taken so that resilvering doesn't complete
30 # before we exported the pool.
31 # 4. Verify that we can import the pool using the first cachefile backup.
32 # (Test 1. cachefile: pre-replace, pool: resilvering)
33 # 5. Wait for the resilvering to finish and export the pool.
34 # 6. Verify that we can import the pool using the first cachefile backup.
35 # (Test 2. cachefile: pre-replace, pool: post-replace)
36 # 7. Export the pool.
37 # 8. Verify that we can import the pool using the second cachefile backup.
38 # (Test 3. cachefile: resilvering, pool: post-replace)
40 # STRATEGY TO SLOW DOWN RESILVERING:
41 # 1. Reduce zfs_txg_timeout, which controls how long can we resilver for
42 # each sync.
43 # 2. Add data to pool
44 # 3. Re-import the pool so that data isn't cached
45 # 4. Use zinject to slow down device I/O
46 # 5. Trigger the resilvering
47 # 6. Use spa freeze to stop writing to the pool.
48 # 7. Clear zinject events (needed to export the pool)
49 # 8. Export the pool
52 verify_runnable "global"
54 ZFS_TXG_TIMEOUT=""
56 function custom_cleanup
58 # Revert zfs_txg_timeout to defaults
59 [[ -n ZFS_TXG_TIMEOUT ]] &&
60 log_must set_zfs_txg_timeout $ZFS_TXG_TIMEOUT
62 cleanup
65 log_onexit custom_cleanup
67 function test_replacing_vdevs
69 typeset poolcreate="$1"
70 typeset replacevdev="$2"
71 typeset replaceby="$3"
72 typeset poolfinalstate="$4"
73 typeset zinjectdevices="$5"
74 typeset earlyremove="$6"
76 log_note "$0: pool '$poolcreate', replace $replacevdev by $replaceby."
78 log_must zpool create -o cachefile=$CPATH $TESTPOOL1 $poolcreate
80 # Cachefile: pool in pre-replace state
81 log_must cp $CPATH $CPATHBKP
83 # Steps to insure resilvering happens very slowly.
84 log_must write_some_data $TESTPOOL1
85 log_must zpool export $TESTPOOL1
86 log_must cp $CPATHBKP $CPATH
87 log_must zpool import -c $CPATH -o cachefile=$CPATH $TESTPOOL1
88 typeset device
89 for device in $zinjectdevices ; do
90 log_must zinject -d $device -D 200:1 $TESTPOOL1 > /dev/null
91 done
92 log_must zpool replace $TESTPOOL1 $replacevdev $replaceby
94 # Cachefile: pool in resilvering state
95 log_must cp $CPATH $CPATHBKP2
97 # We must disable zinject in order to export the pool, so we freeze
98 # it first to prevent writing out subsequent resilvering progress.
99 log_must zpool freeze $TESTPOOL1
100 # Confirm pool is still replacing
101 log_must pool_is_replacing $TESTPOOL1
102 log_must zinject -c all > /dev/null
103 log_must zpool export $TESTPOOL1
105 ( $earlyremove ) && log_must rm $replacevdev
107 ############################################################
108 # Test 1. Cachefile: pre-replace, pool: resilvering
109 ############################################################
110 log_must cp $CPATHBKP $CPATH
111 log_must zpool import -c $CPATH $TESTPOOL1
113 # Wait for resilvering to finish
114 log_must wait_for_pool_config $TESTPOOL1 "$poolfinalstate"
115 log_must zpool export $TESTPOOL1
117 ( ! $earlyremove ) && log_must rm $replacevdev
119 ############################################################
120 # Test 2. Cachefile: pre-replace, pool: post-replace
121 ############################################################
122 log_must zpool import -c $CPATHBKP $TESTPOOL1
123 log_must check_pool_config $TESTPOOL1 "$poolfinalstate"
124 log_must zpool export $TESTPOOL1
126 ############################################################
127 # Test 3. Cachefile: resilvering, pool: post-replace
128 ############################################################
129 log_must zpool import -c $CPATHBKP2 $TESTPOOL1
130 log_must check_pool_config $TESTPOOL1 "$poolfinalstate"
132 # Cleanup
133 log_must zpool destroy $TESTPOOL1
134 log_must rm -f $CPATH $CPATHBKP $CPATHBKP2
135 log_must mkfile $FILE_SIZE $replacevdev
137 log_note ""
140 # We set zfs_txg_timeout to 1 to reduce resilvering time at each sync.
141 ZFS_TXG_TIMEOUT=$(get_zfs_txg_timeout)
142 set_zfs_txg_timeout 1
144 test_replacing_vdevs "$VDEV0 $VDEV1" \
145 "$VDEV1" "$VDEV2" \
146 "$VDEV0 $VDEV2" \
147 "$VDEV0 $VDEV1" \
148 false
150 test_replacing_vdevs "mirror $VDEV0 $VDEV1" \
151 "$VDEV1" "$VDEV2" \
152 "mirror $VDEV0 $VDEV2" \
153 "$VDEV0 $VDEV1" \
154 true
156 test_replacing_vdevs "raidz $VDEV0 $VDEV1 $VDEV2" \
157 "$VDEV1" "$VDEV3" \
158 "raidz $VDEV0 $VDEV3 $VDEV2" \
159 "$VDEV0 $VDEV1 $VDEV2" \
160 true
162 set_zfs_txg_timeout $ZFS_TXG_TIMEOUT
164 log_pass "zpool import -c cachefile_unaware_of_replace passed."