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
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
22 # It should be possible to rewind a pool beyond a device replacement.
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.
37 # STRATEGY TO SLOW DOWN RESILVERING:
38 # 1. Reduce zfs_txg_timeout, which controls how long can we resilver for
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)
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
56 verify_runnable "global"
60 function custom_cleanup
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 SCAN_SUSPEND_PROGRESS 0
70 log_onexit custom_cleanup
72 function test_replace_vdev
74 typeset poolcreate="$1"
75 typeset replacevdev="$2"
76 typeset replaceby="$3"
77 typeset poolfinalstate="$4"
78 typeset zinjectdevices="$5"
79 typeset writedata="$6"
81 log_note "$0: pool '$poolcreate', replace $replacevdev by $replaceby."
83 log_must zpool create $TESTPOOL1 $poolcreate
85 # generate data and checksum it
86 log_must generate_data $TESTPOOL1 $MD5FILE
88 # add more data so that resilver takes longer
89 log_must write_some_data $TESTPOOL1 $writedata
91 # Syncing a few times while writing new data increases the odds that
92 # MOS metadata for some of the txgs will survive.
93 log_must sync_some_data_a_few_times $TESTPOOL1
95 txg=$(get_last_txg_synced $TESTPOOL1)
96 log_must zfs snapshot -r $TESTPOOL1@snap1
98 # This should not free original data.
99 log_must overwrite_data $TESTPOOL1 ""
101 log_must zpool export $TESTPOOL1
102 log_must zpool import -d $DEVICE_DIR $TESTPOOL1
104 # Ensure resilvering doesn't complete.
105 log_must set_tunable32 SCAN_SUSPEND_PROGRESS 1
106 log_must zpool replace $TESTPOOL1 $replacevdev $replaceby
108 # Confirm pool is still replacing
109 log_must pool_is_replacing $TESTPOOL1
110 log_must zpool export $TESTPOOL1
111 log_must set_tunable32 SCAN_SUSPEND_PROGRESS 0
113 ############################################################
114 # Test 1: rewind while device is resilvering.
115 # Import read only to avoid overwriting more recent blocks.
116 ############################################################
117 log_must zpool import -d $DEVICE_DIR -o readonly=on -T $txg $TESTPOOL1
118 log_must check_pool_config $TESTPOOL1 "$poolcreate"
120 log_must verify_data_md5sums $MD5FILE
122 log_must zpool export $TESTPOOL1
124 # Import pool at latest txg to finish the resilvering
125 log_must zpool import -d $DEVICE_DIR $TESTPOOL1
126 log_must overwrite_data $TESTPOOL1 ""
127 log_must wait_for_pool_config $TESTPOOL1 "$poolfinalstate"
128 log_must zpool export $TESTPOOL1
130 # Move out the new device
131 log_must mv $replaceby $BACKUP_DEVICE_DIR/
133 ############################################################
134 # Test 2: rewind after device has been replaced.
135 # Import read-write since we won't need the pool anymore.
136 ############################################################
137 log_must zpool import -d $DEVICE_DIR -T $txg $TESTPOOL1
138 log_must check_pool_config $TESTPOOL1 "$poolcreate"
140 log_must verify_data_md5sums $MD5FILE
143 log_must zpool destroy $TESTPOOL1
144 # Restore the device we moved out
145 log_must mv "$BACKUP_DEVICE_DIR/$(basename $replaceby)" $DEVICE_DIR/
146 # Fast way to clear vdev labels
147 log_must zpool create -f $TESTPOOL2 $VDEV0 $VDEV1 $VDEV2 $VDEV3 $VDEV4
148 log_must zpool destroy $TESTPOOL2
154 is_linux && log_must set_tunable32 TXG_HISTORY 100
156 log_must mkdir -p $BACKUP_DEVICE_DIR
157 # Make the devices bigger to reduce chances of overwriting MOS metadata.
158 increase_device_sizes $(( FILE_SIZE * 4 ))
160 # We set zfs_txg_timeout to 1 to reduce resilvering time at each sync.
161 ZFS_TXG_TIMEOUT=$(get_zfs_txg_timeout)
162 set_zfs_txg_timeout 1
164 test_replace_vdev "$VDEV0 $VDEV1" \
169 test_replace_vdev "mirror $VDEV0 $VDEV1" \
171 "mirror $VDEV0 $VDEV2" \
174 test_replace_vdev "raidz $VDEV0 $VDEV1 $VDEV2" \
176 "raidz $VDEV0 $VDEV3 $VDEV2" \
177 "$VDEV0 $VDEV1 $VDEV2" 10
179 set_zfs_txg_timeout $ZFS_TXG_TIMEOUT
181 log_pass "zpool import rewind after device replacement passed."