xref: /freebsd/sys/contrib/openzfs/tests/zfs-tests/tests/functional/fault/auto_spare_002_pos.ksh (revision 61145dc2b94f12f6a47344fb9aac702321880e43)
1#!/bin/ksh -p
2# SPDX-License-Identifier: CDDL-1.0
3
4#
5# CDDL HEADER START
6#
7# This file and its contents are supplied under the terms of the
8# Common Development and Distribution License ("CDDL"), version 1.0.
9# You may only use this file in accordance with the terms of version
10# 1.0 of the CDDL.
11#
12# A full copy of the text of the CDDL should have accompanied this
13# source.  A copy of the CDDL is also available via the Internet at
14# http://www.illumos.org/license/CDDL.
15#
16# CDDL HEADER END
17#
18
19#
20# Copyright (c) 2017 by Intel Corporation. All rights reserved.
21#
22
23. $STF_SUITE/include/libtest.shlib
24. $STF_SUITE/tests/functional/fault/fault.cfg
25
26#
27# DESCRIPTION:
28# Testing Fault Management Agent ZED Logic - Automated Auto-Spare Test when
29# drive is faulted due to CHECKSUM ERRORS.
30#
31# STRATEGY:
32# 1. Create a pool with hot spares
33# 2. Create a filesystem with the primary cache disable to force reads
34# 3. Write a file to the pool to be read back
35# 4. Inject CHECKSUM ERRORS on read with a zinject error handler
36# 5. Verify the ZED kicks in a hot spare and expected pool/device status
37# 6. Clear the fault
38# 7. Verify the hot spare is available and expected pool/device status
39#
40
41verify_runnable "both"
42
43function cleanup
44{
45	log_must zinject -c all
46	destroy_pool $TESTPOOL
47	rm -f $VDEV_FILES $SPARE_FILE
48}
49
50log_assert "Testing automated auto-spare FMA test"
51
52log_onexit cleanup
53
54# Events not supported on FreeBSD
55if ! is_freebsd; then
56	# Clear events from previous runs
57	zed_events_drain
58fi
59
60TESTFILE="/$TESTPOOL/$TESTFS/testfile"
61
62for type in "mirror" "raidz" "raidz2"; do
63	# 1. Create a pool with hot spares
64	log_must truncate -s $MINVDEVSIZE $VDEV_FILES $SPARE_FILE
65	log_must zpool create -f $TESTPOOL $type $VDEV_FILES \
66	    spare $SPARE_FILE
67
68	# 2. Create a filesystem with the primary cache disable to force reads
69	log_must zfs create -o primarycache=none $TESTPOOL/$TESTFS
70	log_must zfs set recordsize=16k $TESTPOOL/$TESTFS
71
72	# 3. Write a file to the pool to be read back
73	log_must dd if=/dev/urandom of=$TESTFILE bs=1M count=64
74
75	# 4. Inject CHECKSUM ERRORS on read with a zinject error handler
76	log_must zinject -d $FAULT_FILE -e corrupt -f 50 -T read $TESTPOOL
77	log_must dd if=$TESTFILE of=/dev/null bs=1M count=64
78
79	# 5. Verify the ZED kicks in a hot spare and expected pool/device status
80	log_note "Wait for ZED to auto-spare"
81	log_must wait_vdev_state $TESTPOOL $FAULT_FILE "DEGRADED" 60
82	log_must wait_vdev_state $TESTPOOL $SPARE_FILE "ONLINE" 60
83	log_must wait_hotspare_state $TESTPOOL $SPARE_FILE "INUSE"
84	log_must check_state $TESTPOOL "" "DEGRADED"
85
86	# 6. Clear the fault
87	log_must zinject -c all
88	log_must zpool clear $TESTPOOL $FAULT_FILE
89
90	# 7. Verify the hot spare is available and expected pool/device status
91	log_must wait_vdev_state $TESTPOOL $FAULT_FILE "ONLINE" 60
92	log_must wait_hotspare_state $TESTPOOL $SPARE_FILE "AVAIL"
93	log_must check_state $TESTPOOL "" "ONLINE"
94
95	cleanup
96done
97
98log_pass "Auto-spare test successful"
99