vdev_ashift should only be set once

== Motivation and Context

The new vdev ashift optimization prevents the removal of devices when
a zfs configuration is comprised of disks which have different logical
and physical block sizes. This is caused because we set 'spa_min_ashift'
in vdev_open and then later call 'vdev_ashift_optimize'. This would
result in an inconsistency between spa's ashift calculations and that
of the top-level vdev.

In addition, the optimization logical ignores the overridden ashift
value that would be provided by '-o ashift=<val>'.

== Description

This change reworks the vdev ashift optimization so that it's only
set the first time the device is configured. It still allows the
physical and logical ahsift values to be set every time the device
is opened but those values are only consulted on first open.

Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Cedric Berger <cedric@precidata.com>
Signed-off-by: George Wilson <gwilson@delphix.com>
External-Issue: DLPX-71831
Closes #10932
This commit is contained in:
George Wilson
2020-09-18 14:13:47 -05:00
committed by GitHub
parent 908d43d0a9
commit c494aa7f57
19 changed files with 198 additions and 52 deletions
-2
View File
@@ -9434,8 +9434,6 @@ l2arc_add_vdev(spa_t *spa, vdev_t *vd)
ASSERT(!l2arc_vdev_present(vd));
vdev_ashift_optimize(vd);
/*
* Create a new l2arc device entry.
*/
-1
View File
@@ -5762,7 +5762,6 @@ spa_create(const char *pool, nvlist_t *nvroot, nvlist_t *props,
for (int c = 0; error == 0 && c < rvd->vdev_children; c++) {
vdev_t *vd = rvd->vdev_child[c];
vdev_ashift_optimize(vd);
vdev_metaslab_set_size(vd);
vdev_expand(vd, txg);
}
+1 -3
View File
@@ -577,10 +577,8 @@ spa_config_update(spa_t *spa, int what)
(tvd->vdev_islog && tvd->vdev_removing))
continue;
if (tvd->vdev_ms_array == 0) {
vdev_ashift_optimize(tvd);
if (tvd->vdev_ms_array == 0)
vdev_metaslab_set_size(tvd);
}
vdev_expand(tvd, txg);
}
}
+59 -37
View File
@@ -1672,6 +1672,38 @@ vdev_set_deflate_ratio(vdev_t *vd)
}
}
/*
* Maximize performance by inflating the configured ashift for top level
* vdevs to be as close to the physical ashift as possible while maintaining
* administrator defined limits and ensuring it doesn't go below the
* logical ashift.
*/
static void
vdev_ashift_optimize(vdev_t *vd)
{
ASSERT(vd == vd->vdev_top);
if (vd->vdev_ashift < vd->vdev_physical_ashift) {
vd->vdev_ashift = MIN(
MAX(zfs_vdev_max_auto_ashift, vd->vdev_ashift),
MAX(zfs_vdev_min_auto_ashift,
vd->vdev_physical_ashift));
} else {
/*
* If the logical and physical ashifts are the same, then
* we ensure that the top-level vdev's ashift is not smaller
* than our minimum ashift value. For the unusual case
* where logical ashift > physical ashift, we can't cap
* the calculated ashift based on max ashift as that
* would cause failures.
* We still check if we need to increase it to match
* the min ashift.
*/
vd->vdev_ashift = MAX(zfs_vdev_min_auto_ashift,
vd->vdev_ashift);
}
}
/*
* Prepare a virtual device for access.
*/
@@ -1830,16 +1862,17 @@ vdev_open(vdev_t *vd)
return (SET_ERROR(EINVAL));
}
/*
* We can always set the logical/physical ashift members since
* their values are only used to calculate the vdev_ashift when
* the device is first added to the config. These values should
* not be used for anything else since they may change whenever
* the device is reopened and we don't store them in the label.
*/
vd->vdev_physical_ashift =
MAX(physical_ashift, vd->vdev_physical_ashift);
vd->vdev_logical_ashift = MAX(logical_ashift, vd->vdev_logical_ashift);
vd->vdev_ashift = MAX(vd->vdev_logical_ashift, vd->vdev_ashift);
if (vd->vdev_logical_ashift > ASHIFT_MAX) {
vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_ASHIFT_TOO_BIG);
return (SET_ERROR(EDOM));
}
vd->vdev_logical_ashift = MAX(logical_ashift,
vd->vdev_logical_ashift);
if (vd->vdev_asize == 0) {
/*
@@ -1848,6 +1881,24 @@ vdev_open(vdev_t *vd)
*/
vd->vdev_asize = asize;
vd->vdev_max_asize = max_asize;
/*
* If the vdev_ashift was not overriden at creation time,
* then set it the logical ashift and optimize the ashift.
*/
if (vd->vdev_ashift == 0) {
vd->vdev_ashift = vd->vdev_logical_ashift;
if (vd->vdev_logical_ashift > ASHIFT_MAX) {
vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
VDEV_AUX_ASHIFT_TOO_BIG);
return (SET_ERROR(EDOM));
}
if (vd->vdev_top == vd) {
vdev_ashift_optimize(vd);
}
}
if (vd->vdev_ashift != 0 && (vd->vdev_ashift < ASHIFT_MIN ||
vd->vdev_ashift > ASHIFT_MAX)) {
vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
@@ -2444,35 +2495,6 @@ vdev_metaslab_set_size(vdev_t *vd)
ASSERT3U(vd->vdev_ms_shift, >=, SPA_MAXBLOCKSHIFT);
}
/*
* Maximize performance by inflating the configured ashift for top level
* vdevs to be as close to the physical ashift as possible while maintaining
* administrator defined limits and ensuring it doesn't go below the
* logical ashift.
*/
void
vdev_ashift_optimize(vdev_t *vd)
{
if (vd == vd->vdev_top) {
if (vd->vdev_ashift < vd->vdev_physical_ashift) {
vd->vdev_ashift = MIN(
MAX(zfs_vdev_max_auto_ashift, vd->vdev_ashift),
MAX(zfs_vdev_min_auto_ashift,
vd->vdev_physical_ashift));
} else {
/*
* Unusual case where logical ashift > physical ashift
* so we can't cap the calculated ashift based on max
* ashift as that would cause failures.
* We still check if we need to increase it to match
* the min ashift.
*/
vd->vdev_ashift = MAX(zfs_vdev_min_auto_ashift,
vd->vdev_ashift);
}
}
}
void
vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg)
{
+1 -1
View File
@@ -391,7 +391,7 @@ vdev_mirror_open(vdev_t *vd, uint64_t *asize, uint64_t *max_asize,
*max_asize = MIN(*max_asize - 1, cvd->vdev_max_asize - 1) + 1;
*logical_ashift = MAX(*logical_ashift, cvd->vdev_ashift);
*physical_ashift = MAX(*physical_ashift,
vd->vdev_physical_ashift);
cvd->vdev_physical_ashift);
}
if (numerrors == vd->vdev_children) {