Skip to content

Commit

Permalink
Improve tests and update man page for healing recv
Browse files Browse the repository at this point in the history
Fix the manpage. The "SYNOPSIS" section is incorrectly formatted for 
receive -c.  I also took this opportunity to reword some parts and 
fix a run-on sentence in the manpage.

Add large block testing for corrective recv. This adds a new test 
that makes sure blocks generated using zfs send -L/--large-block 
large-block send flag are able to be used for healing.

Since with unloaded key and errlog feature enabled corruption is not 
shown in zpool status #13675 is fixed the zfs_receive_corrective.ksh 
test no longer sets -o feature@head_errlog=disabled on pool creation 
so that it can also test for regressions related to head_errlog feature.

Note that the zfs_receive_compressed_corrective.ksh and
zfs_receive_large_block_corrective.ksh tests are still creating pools 
with -o feature@head_errlog=disabled.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Alek Pinchuk <apinchuk@axcient.com>
Closes #14615
  • Loading branch information
alek-p authored Mar 15, 2023
1 parent d74f123 commit f55d6ee
Show file tree
Hide file tree
Showing 5 changed files with 204 additions and 10 deletions.
15 changes: 7 additions & 8 deletions man/man8/zfs-receive.8
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@
.\" Copyright 2018 Nexenta Systems, Inc.
.\" Copyright 2019 Joyent, Inc.
.\"
.Dd April 26, 2022
.Dd March 12, 2023
.Dt ZFS-RECEIVE 8
.Os
.
Expand All @@ -56,8 +56,7 @@
.Cm receive
.Fl A
.Ar filesystem Ns | Ns Ar volume
.
.Nm
.Nm zfs
.Cm receive
.Fl c
.Op Fl vn
Expand Down Expand Up @@ -406,15 +405,15 @@ deleting its saved partially received state.
.Op Fl vn
.Ar filesystem Ns | Ns Ar snapshot
.Xc
Attempt to correct data corruption in the specified dataset,
Attempt to repair data corruption in the specified dataset,
by using the provided stream as the source of healthy data.
This method of healing can only heal data blocks present in the stream.
Metadata can not be healed by corrective receive.
Running a scrub is recommended post-healing to ensure all corruption was
healed.
Running a scrub is recommended post-healing to ensure all data corruption was
repaired.
.Pp
It's important to consider why corruption has happened in the first place
since if you have slowly failing hardware periodically healing the data
It's important to consider why corruption has happened in the first place.
If you have slowly failing hardware - periodically repairing the data
is not going to save you from data loss later on when the hardware fails
completely.
.El
Expand Down
2 changes: 1 addition & 1 deletion tests/runfiles/common.run
Original file line number Diff line number Diff line change
Expand Up @@ -242,7 +242,7 @@ tests = ['zfs_receive_001_pos', 'zfs_receive_002_pos', 'zfs_receive_003_pos',
'zfs_receive_raw', 'zfs_receive_raw_incremental', 'zfs_receive_-e',
'zfs_receive_raw_-d', 'zfs_receive_from_zstd', 'zfs_receive_new_props',
'zfs_receive_-wR-encrypted-mix', 'zfs_receive_corrective',
'zfs_receive_compressed_corrective']
'zfs_receive_compressed_corrective', 'zfs_receive_large_block_corrective']
tags = ['functional', 'cli_root', 'zfs_receive']

[tests/functional/cli_root/zfs_rename]
Expand Down
1 change: 1 addition & 0 deletions tests/zfs-tests/tests/Makefile.am
Original file line number Diff line number Diff line change
Expand Up @@ -774,6 +774,7 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
functional/cli_root/zfs_receive/zfs_receive_-wR-encrypted-mix.ksh \
functional/cli_root/zfs_receive/zfs_receive_corrective.ksh \
functional/cli_root/zfs_receive/zfs_receive_compressed_corrective.ksh \
functional/cli_root/zfs_receive/zfs_receive_large_block_corrective.ksh \
functional/cli_root/zfs_rename/cleanup.ksh \
functional/cli_root/zfs_rename/setup.ksh \
functional/cli_root/zfs_rename/zfs_rename_001_pos.ksh \
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -85,7 +85,7 @@ typeset passphrase="password"
typeset file="/$TESTPOOL/$TESTFS1/$TESTFILE0"

log_must eval "poolexists $TESTPOOL && destroy_pool $TESTPOOL"
log_must zpool create -f -o feature@head_errlog=disabled $TESTPOOL $DISK
log_must zpool create -f $TESTPOOL $DISK

log_must eval "echo $passphrase > /$TESTPOOL/pwd"

Expand Down
Original file line number Diff line number Diff line change
@@ -0,0 +1,194 @@
#!/bin/ksh -p
#
# CDDL HEADER START
#
# This file and its contents are supplied under the terms of the
# Common Development and Distribution License ("CDDL"), version 1.0.
# You may only use this file in accordance with the terms of version
# 1.0 of the CDDL.
#
# A full copy of the text of the CDDL should have accompanied this
# source. A copy of the CDDL is also available via the Internet at
# http://www.illumos.org/license/CDDL.
#
# CDDL HEADER END
#

#
# Copyright (c) 2019 Datto, Inc. All rights reserved.
# Copyright (c) 2022 Axcient.
#

. $STF_SUITE/include/libtest.shlib

#
# DESCRIPTION:
# OpenZFS should be able to heal data using corrective recv when the send file
# was generated with the --large-block flag
#
# STRATEGY:
# 0. Create a file, checksum the file to be corrupted then compare it's checksum
# with the one obtained after healing under different testing scenarios:
# 1. Test healing (aka corrective) recv from a full send file
# 2. Test healing recv (aka heal recv) from an incremental send file
# 3. Test healing recv when compression on-disk is off but source was compressed
# 4. Test heal recv when compression on-disk is on but source was uncompressed
# 5. Test heal recv when compression doesn't match between send file and on-disk
# 6. Test healing recv of an encrypted dataset using an unencrypted send file
# 7. Test healing recv (on an encrypted dataset) using a raw send file
# 8. Test healing when specifying destination filesystem only (no snapshot)
# 9. Test incremental recv aftear healing recv
#

verify_runnable "both"

DISK=${DISKS%% *}

backup=$TEST_BASE_DIR/backup
raw_backup=$TEST_BASE_DIR/raw_backup
ibackup=$TEST_BASE_DIR/ibackup
unc_backup=$TEST_BASE_DIR/unc_backup

function cleanup
{
log_must rm -f $backup $raw_backup $ibackup $unc_backup

poolexists $TESTPOOL && destroy_pool $TESTPOOL
log_must zpool create -f $TESTPOOL $DISK
}

function test_corrective_recv
{
log_must zpool scrub -w $TESTPOOL
log_must zpool status -v $TESTPOOL
log_must eval "zpool status -v $TESTPOOL | \
grep \"Permanent errors have been detected\""

# make sure we will read the corruption from disk by flushing the ARC
log_must zinject -a

log_must eval "zfs recv -c $1 < $2"

log_must zpool scrub -w $TESTPOOL
log_must zpool status -v $TESTPOOL
log_mustnot eval "zpool status -v $TESTPOOL | \
grep \"Permanent errors have been detected\""
typeset cksum=$(md5digest $file)
[[ "$cksum" == "$checksum" ]] || \
log_fail "Checksums differ ($cksum != $checksum)"
}

log_onexit cleanup

log_assert "ZFS corrective receive should be able to heal data corruption"

typeset passphrase="password"
typeset file="/$TESTPOOL/$TESTFS1/$TESTFILE0"

log_must eval "poolexists $TESTPOOL && destroy_pool $TESTPOOL"
log_must zpool create -f -o feature@large_blocks=enabled \
-o feature@head_errlog=disabled $TESTPOOL $DISK

log_must eval "echo $passphrase > /$TESTPOOL/pwd"

log_must zfs create -o recordsize=1m -o primarycache=none \
-o atime=off -o compression=lz4 $TESTPOOL/$TESTFS1

log_must dd if=/dev/urandom of=$file bs=1024 count=1024 oflag=sync
log_must eval "echo 'aaaaaaaa' >> "$file
typeset checksum=$(md5digest $file)

log_must zfs snapshot $TESTPOOL/$TESTFS1@snap1

# create full send file
log_must eval "zfs send -L $TESTPOOL/$TESTFS1@snap1 > $backup"

log_must dd if=/dev/urandom of=$file"1" bs=1024 count=1024 oflag=sync
log_must eval "echo 'bbbbbbbb' >> "$file"1"
log_must zfs snapshot $TESTPOOL/$TESTFS1@snap2
# create incremental send file
log_must eval "zfs send -Li $TESTPOOL/$TESTFS1@snap1 \
$TESTPOOL/$TESTFS1@snap2 > $ibackup"

corrupt_blocks_at_level $file 0
# test healing recv from a full send file
test_corrective_recv $TESTPOOL/$TESTFS1@snap1 $backup

corrupt_blocks_at_level $file"1" 0
# test healing recv from an incremental send file
test_corrective_recv $TESTPOOL/$TESTFS1@snap2 $ibackup

# create new uncompressed dataset using our send file
log_must eval "zfs recv -o compression=off -o primarycache=none \
$TESTPOOL/$TESTFS2 < $backup"
typeset compr=$(get_prop compression $TESTPOOL/$TESTFS2)
[[ "$compr" == "off" ]] || \
log_fail "Unexpected compression $compr in recved dataset"
corrupt_blocks_at_level "/$TESTPOOL/$TESTFS2/$TESTFILE0" 0
# test healing recv when compression on-disk is off but source was compressed
test_corrective_recv "$TESTPOOL/$TESTFS2@snap1" $backup

# create a full sendfile from an uncompressed source
log_must eval "zfs send -L $TESTPOOL/$TESTFS2@snap1 > $unc_backup"
log_must eval "zfs recv -o compression=gzip -o primarycache=none \
-o recordsize=1m $TESTPOOL/testfs3 < $unc_backup"
typeset compr=$(get_prop compression $TESTPOOL/testfs3)
[[ "$compr" == "gzip" ]] || \
log_fail "Unexpected compression $compr in recved dataset"
corrupt_blocks_at_level "/$TESTPOOL/testfs3/$TESTFILE0" 0
# test healing recv when compression on-disk is on but source was uncompressed
test_corrective_recv "$TESTPOOL/testfs3@snap1" $unc_backup

# create new compressed dataset using our send file
log_must eval "zfs recv -o compression=gzip -o primarycache=none \
-o recordsize=1m $TESTPOOL/testfs4 < $backup"
typeset compr=$(get_prop compression $TESTPOOL/testfs4)
[[ "$compr" == "gzip" ]] || \
log_fail "Unexpected compression $compr in recved dataset"
corrupt_blocks_at_level "/$TESTPOOL/testfs4/$TESTFILE0" 0
# test healing recv when compression doesn't match between send file and on-disk
test_corrective_recv "$TESTPOOL/testfs4@snap1" $backup

# create new encrypted (and compressed) dataset using our send file
log_must eval "zfs recv -o encryption=aes-256-ccm -o keyformat=passphrase \
-o recordsize=1m -o keylocation=file:///$TESTPOOL/pwd -o primarycache=none \
$TESTPOOL/testfs5 < $backup"
typeset encr=$(get_prop encryption $TESTPOOL/testfs5)
[[ "$encr" == "aes-256-ccm" ]] || \
log_fail "Unexpected encryption $encr in recved dataset"
log_must eval "zfs send -L --raw $TESTPOOL/testfs5@snap1 > $raw_backup"
log_must eval "zfs send -L $TESTPOOL/testfs5@snap1 > $backup"
corrupt_blocks_at_level "/$TESTPOOL/testfs5/$TESTFILE0" 0
# test healing recv of an encrypted dataset using an unencrypted send file
test_corrective_recv "$TESTPOOL/testfs5@snap1" $backup
corrupt_blocks_at_level "/$TESTPOOL/testfs5/$TESTFILE0" 0
log_must zfs unmount $TESTPOOL/testfs5
log_must zfs unload-key $TESTPOOL/testfs5
# test healing recv (on an encrypted dataset) using a raw send file
test_corrective_recv "$TESTPOOL/testfs5@snap1" $raw_backup
# non raw send file healing an encrypted dataset with an unloaded key will fail
log_mustnot eval "zfs recv -c $TESTPOOL/testfs5@snap1 < $backup"

log_must zfs rollback -r $TESTPOOL/$TESTFS1@snap1
corrupt_blocks_at_level $file 0
# test healing when specifying destination filesystem only (no snapshot)
test_corrective_recv $TESTPOOL/$TESTFS1 $backup
# test incremental recv aftear healing recv
log_must eval "zfs recv -o recordsize=1m $TESTPOOL/$TESTFS1 < $ibackup"

# test that healing recv can not be combined with incompatible recv options
log_mustnot eval "zfs recv -h -c $TESTPOOL/$TESTFS1@snap1 < $backup"
log_mustnot eval "zfs recv -F -c $TESTPOOL/$TESTFS1@snap1 < $backup"
log_mustnot eval "zfs recv -s -c $TESTPOOL/$TESTFS1@snap1 < $backup"
log_mustnot eval "zfs recv -u -c $TESTPOOL/$TESTFS1@snap1 < $backup"
log_mustnot eval "zfs recv -d -c $TESTPOOL/$TESTFS1@snap1 < $backup"
log_mustnot eval "zfs recv -e -c $TESTPOOL/$TESTFS1@snap1 < $backup"

# ensure healing recv doesn't work when snap GUIDS don't match
log_mustnot eval "zfs recv -c $TESTPOOL/testfs5@snap2 < $backup"
log_mustnot eval "zfs recv -c $TESTPOOL/testfs5 < $backup"

# test that healing recv doesn't work on non-existing snapshots
log_mustnot eval "zfs recv -c $TESTPOOL/$TESTFS1@missing < $backup"

log_pass "OpenZFS corrective recv works for data healing"

0 comments on commit f55d6ee

Please sign in to comment.