I sent this out almost two weeks ago, but I haven't heard anything back
about whether this behavior is intentional or not. I suspect a bunch of
people may have been out around that time due to the US July 4 holiday. I
thought I would post the small patch I came up with to see if anyone has
any feedback. If I could get a definitive answer that this behavior is a
bug, I'll open a bug report and work on getting a patch integrated.
--- a/usr/src/uts/common/fs/zfs/spa.c
+++ b/usr/src/uts/common/fs/zfs/spa.c
@@ -6535,6 +6535,7 @@ spa_vdev_resilver_done_hunt(vdev_t *vd)
/*
* Check for a completed resilver with the 'unspare' flag set.
+ * Also potentially update faulted state.
*/
if (vd->vdev_ops == &vdev_spare_ops) {
vdev_t *first = vd->vdev_child[0];
@@ -6557,6 +6558,26 @@ spa_vdev_resilver_done_hunt(vdev_t *vd)
return (oldvd);
/*
+ * We know the vdev is a spare (e.g. "spare-1") which just
+ * finished resilvering. If it's faulted, and one of the
+ * children is healthy, then set the spare's state to
degraded
+ * so that it will handle read operations.
+ */
+ if (vd->vdev_state == VDEV_STATE_FAULTED &&
+ vd->vdev_children >= 2) {
+ int i;
+
+ for (i = 0; i < vd->vdev_children; i++) {
+ if (vd->vdev_child[i]->vdev_state ==
+ VDEV_STATE_HEALTHY) {
+ vdev_set_state(vd, B_FALSE,
+ VDEV_STATE_DEGRADED,
VDEV_AUX_NONE);
+ break;
+ }
+ }
+ }
+
+ /*
* If there are more than two spares attached to a disk,
* and those spares are not required, then we want to
* attempt to free them up now so that they can be used
Thanks,
Jerry
On Thu, Jul 5, 2018 at 10:29 AM, Jerry Jelinek <[email protected]>
wrote:
> Is it intentional that ZFS does not read from a hot spare under a mirror,
> even after the resilver has completed?
>
> We have observed this behavior and I spent some time digging into it. Here
> is an example of the zpool status once the resilver has completed:
>
> NAME STATE READ WRITE CKSUM
> jjpool DEGRADED 0 0 0
> mirror-0 DEGRADED 0 0 0
> c1t2d0 ONLINE 0 0 0
> spare-1 FAULTED 0 0 0
> c1t3d0 FAULTED 0 0 0 too many errors
> c1t4d0 ONLINE 0 0 0
> spares
> c1t4d0 INUSE currently in use
>
> Because the "spare-1" vdev stays in the "faulted" state, vdev_readable()
> always returns false for that vdev and all of the read workload is going to
> the other side of the mirror.
>
> I've tried to determine if this is a regression, intentional, or just an
> oversight, but I am not sure.
>
> If this is an oversight, I have a small change which will change the
> "spare" vdev state to "degraded" once the hot-spare resilver has completed.
> This allows reads to go to the healthy hot-spare under the "spare-1" vdev.
> If that seems reasonable, I could get the patch ready for review.
>
> Thanks,
> Jerry
>
>
------------------------------------------
openzfs: openzfs-developer
Permalink:
https://openzfs.topicbox.com/groups/developer/Tc7b699f05a31ea88-Mb2c1261756537f66dfadfc13
Delivery options: https://openzfs.topicbox.com/groups
