c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2026-05-23 02:44:41 +03:00
Code Issues Packages Projects Releases Wiki Activity

Restrict raidz faulted vdev count

Browse Source 
Specifically, a child in a replacing vdev won't count when assessing
the dtl during a vdev_fault()

Sponsored-by: Klara, Inc.
Sponsored-by: Wasabi Technology, Inc.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Tino Reichardt <milky-zfs@mcmilk.de>
Signed-off-by: Don Brady <don.brady@klarasystems.com>
Closes #16569
This commit is contained in:
Don Brady
2024-10-01 10:12:11 -06:00
committed by GitHub
parent 7e957fde73
commit 141368a4b6
4 changed files with 132 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
module/zfs/vdev.c
+33 -10
View File
@@ -3149,9 +3149,9 @@ vdev_dtl_should_excise(vdev_t *vd, boolean_t rebuild_done)
* Reassess DTLs after a config change or scrub completion. If txg == 0 no
* write operations will be issued to the pool.
*/
void
vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg,
boolean_t scrub_done, boolean_t rebuild_done)
static void
vdev_dtl_reassess_impl(vdev_t *vd, uint64_t txg, uint64_t scrub_txg,
boolean_t scrub_done, boolean_t rebuild_done, boolean_t faulting)
{
spa_t *spa = vd->vdev_spa;
avl_tree_t reftree;
@@ -3160,8 +3160,8 @@ vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg,
ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);
for (int c = 0; c < vd->vdev_children; c++)
vdev_dtl_reassess(vd->vdev_child[c], txg,
scrub_txg, scrub_done, rebuild_done);
vdev_dtl_reassess_impl(vd->vdev_child[c], txg,
scrub_txg, scrub_done, rebuild_done, faulting);
if (vd == spa->spa_root_vdev || !vdev_is_concrete(vd) || vd->vdev_aux)
return;
@@ -3255,11 +3255,21 @@ vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg,
if (scrub_done)
range_tree_vacate(vd->vdev_dtl[DTL_SCRUB], NULL, NULL);
range_tree_vacate(vd->vdev_dtl[DTL_OUTAGE], NULL, NULL);
if (!vdev_readable(vd))
/*
* For the faulting case, treat members of a replacing vdev
* as if they are not available. It's more likely than not that
* a vdev in a replacing vdev could encounter read errors so
* treat it as not being able to contribute.
*/
if (!vdev_readable(vd) ||
(faulting && vd->vdev_parent != NULL &&
vd->vdev_parent->vdev_ops == &vdev_replacing_ops)) {
range_tree_add(vd->vdev_dtl[DTL_OUTAGE], 0, -1ULL);
else
} else {
range_tree_walk(vd->vdev_dtl[DTL_MISSING],
range_tree_add, vd->vdev_dtl[DTL_OUTAGE]);
}
/*
* If the vdev was resilvering or rebuilding and no longer
@@ -3321,6 +3331,14 @@ vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg,
}
}
void
vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg,
boolean_t scrub_done, boolean_t rebuild_done)
{
return (vdev_dtl_reassess_impl(vd, txg, scrub_txg, scrub_done,
rebuild_done, B_FALSE));
}
/*
* Iterate over all the vdevs except spare, and post kobj events
*/
@@ -3548,7 +3566,11 @@ vdev_dtl_sync(vdev_t *vd, uint64_t txg)
}
/*
* Determine whether the specified vdev can be offlined/detached/removed
* Determine whether the specified vdev can be
* - offlined
* - detached
* - removed
* - faulted
* without losing data.
*/
boolean_t
@@ -3558,6 +3580,7 @@ vdev_dtl_required(vdev_t *vd)
vdev_t *tvd = vd->vdev_top;
uint8_t cant_read = vd->vdev_cant_read;
boolean_t required;
boolean_t faulting = vd->vdev_state == VDEV_STATE_FAULTED;
ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
@@ -3570,10 +3593,10 @@ vdev_dtl_required(vdev_t *vd)
* If not, we can safely offline/detach/remove the device.
*/
vd->vdev_cant_read = B_TRUE;
vdev_dtl_reassess(tvd, 0, 0, B_FALSE, B_FALSE);
vdev_dtl_reassess_impl(tvd, 0, 0, B_FALSE, B_FALSE, faulting);
required = !vdev_dtl_empty(tvd, DTL_OUTAGE);
vd->vdev_cant_read = cant_read;
vdev_dtl_reassess(tvd, 0, 0, B_FALSE, B_FALSE);
vdev_dtl_reassess_impl(tvd, 0, 0, B_FALSE, B_FALSE, faulting);
if (!required && zio_injection_enabled) {
required = !!zio_handle_device_injection(vd, NULL,