zfs_retire.c revision 88ecc943b4eb72f7c4fbbd8435997b85ef171fc3
99653d4ee642c6528e88224f12409a5f23060994eschrock * CDDL HEADER START
99653d4ee642c6528e88224f12409a5f23060994eschrock * The contents of this file are subject to the terms of the
99653d4ee642c6528e88224f12409a5f23060994eschrock * Common Development and Distribution License (the "License").
99653d4ee642c6528e88224f12409a5f23060994eschrock * You may not use this file except in compliance with the License.
99653d4ee642c6528e88224f12409a5f23060994eschrock * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
99653d4ee642c6528e88224f12409a5f23060994eschrock * See the License for the specific language governing permissions
99653d4ee642c6528e88224f12409a5f23060994eschrock * and limitations under the License.
99653d4ee642c6528e88224f12409a5f23060994eschrock * When distributing Covered Code, include this CDDL HEADER in each
99653d4ee642c6528e88224f12409a5f23060994eschrock * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
99653d4ee642c6528e88224f12409a5f23060994eschrock * If applicable, add the following below this CDDL HEADER, with the
99653d4ee642c6528e88224f12409a5f23060994eschrock * fields enclosed by brackets "[]" replaced with your own identifying
99653d4ee642c6528e88224f12409a5f23060994eschrock * information: Portions Copyright [yyyy] [name of copyright owner]
99653d4ee642c6528e88224f12409a5f23060994eschrock * CDDL HEADER END
cbf75e67acb6c32a2f4884f28a839d59f7988d37Stephen Hanson * Copyright 2009 Sun Microsystems, Inc. All rights reserved.
99653d4ee642c6528e88224f12409a5f23060994eschrock * Use is subject to license terms.
99653d4ee642c6528e88224f12409a5f23060994eschrock * The ZFS retire agent is responsible for managing hot spares across all pools.
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * When we see a device fault or a device removal, we try to open the associated
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * pool and look for any hot spares. We iterate over any available hot spares
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * and attempt a 'zpool replace' for each one.
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * For vdevs diagnosed as faulty, the agent is also responsible for proactively
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * marking the vdev FAULTY (for I/O errors) or DEGRADED (for checksum errors).
99653d4ee642c6528e88224f12409a5f23060994eschrock * Find a pool with a matching GUID.
99653d4ee642c6528e88224f12409a5f23060994eschrocktypedef struct find_cbdata {
99653d4ee642c6528e88224f12409a5f23060994eschrock return (1);
99653d4ee642c6528e88224f12409a5f23060994eschrock return (0);
99653d4ee642c6528e88224f12409a5f23060994eschrock * Find a vdev within a tree with a matching GUID.
99653d4ee642c6528e88224f12409a5f23060994eschrock if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0 &&
99653d4ee642c6528e88224f12409a5f23060994eschrock if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
99653d4ee642c6528e88224f12409a5f23060994eschrock for (c = 0; c < children; c++) {
c5904d138f3bdf0762dbf452a43d5a5c387ea6a8eschrock if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
c5904d138f3bdf0762dbf452a43d5a5c387ea6a8eschrock for (c = 0; c < children; c++) {
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * Given a (pool, vdev) GUID pair, find the matching pool and vdev.
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrockfind_by_guid(libzfs_handle_t *zhdl, uint64_t pool_guid, uint64_t vdev_guid,
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * Find the corresponding pool and make sure the vdev still exists.
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock if ((*vdevp = find_vdev(nvroot, vdev_guid)) == NULL) {
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * Given a vdev, attempt to replace it with every known spare until one
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * succeeds.
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrockreplace_with_spare(zpool_handle_t *zhp, nvlist_t *vdev)
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * Find out if there are any hot spares available in the pool.
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock if (nvlist_alloc(&replacement, NV_UNIQUE_NAME, 0) != 0)
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock if (nvlist_add_string(replacement, ZPOOL_CONFIG_TYPE,
88ecc943b4eb72f7c4fbbd8435997b85ef171fc3George Wilson dev_name = zpool_vdev_name(NULL, zhp, vdev, B_FALSE);
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * Try to replace each spare, ending when we successfully
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * replace it.
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock for (s = 0; s < nspares; s++) {
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH,
99653d4ee642c6528e88224f12409a5f23060994eschrock/*ARGSUSED*/
99653d4ee642c6528e88224f12409a5f23060994eschrockzfs_retire_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl,
99653d4ee642c6528e88224f12409a5f23060994eschrock const char *class)
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * If this is a resource notifying us of device removal, then simply
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * check for an available spare and continue.
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock if (strcmp(class, "resource.fs.zfs.removed") == 0) {
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock if (nvlist_lookup_uint64(nvl, FM_EREPORT_PAYLOAD_ZFS_POOL_GUID,
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock nvlist_lookup_uint64(nvl, FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID,
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock if ((zhp = find_by_guid(zhdl, pool_guid, vdev_guid,
cbf75e67acb6c32a2f4884f28a839d59f7988d37Stephen Hanson if (strcmp(class, FM_LIST_RESOLVED_CLASS) == 0)
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * We subscribe to zfs faults as well as all repair events.
99653d4ee642c6528e88224f12409a5f23060994eschrock if (nvlist_lookup_nvlist_array(nvl, FM_SUSPECT_FAULT_LIST,
99653d4ee642c6528e88224f12409a5f23060994eschrock for (f = 0; f < nfaults; f++) {
cbf75e67acb6c32a2f4884f28a839d59f7988d37Stephen Hanson if (nvlist_lookup_boolean_value(fault, FM_SUSPECT_RETIRE,
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * While we subscribe to fault.fs.zfs.*, we only take action
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * for faults targeting a specific vdev (open failure or SERD
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * failure).
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock if (fmd_nvl_class_match(hdl, fault, "fault.fs.zfs.vdev.io"))
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock else if (fmd_nvl_class_match(hdl, fault, "fault.fs.zfs.device"))
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock if (nvlist_lookup_uint64(resource, FM_FMRI_ZFS_POOL,
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock if ((zhp = find_by_guid(zhdl, pool_guid, vdev_guid,
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * If this is a repair event, then mark the vdev as repaired and
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * continue.
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * Actively fault the device if needed.
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock * Attempt to substitute a hot spare.
25c6ff4b77fcddf4097ce78a8277275ca603b46cstephh if (strcmp(class, FM_LIST_REPAIRED_CLASS) == 0 && repair_done &&
25c6ff4b77fcddf4097ce78a8277275ca603b46cstephh nvlist_lookup_string(nvl, FM_SUSPECT_UUID, &uuid) == 0)