* master - Revert "lvconvert: fix (automatic) raid repair regression"
@ 2016-09-21 15:06 Alasdair Kergon
0 siblings, 0 replies; only message in thread
From: Alasdair Kergon @ 2016-09-21 15:06 UTC (permalink / raw)
To: lvm-devel
Gitweb: http://git.fedorahosted.org/git/?p=lvm2.git;a=commitdiff;h=c2f65806f62a49f6f3da59e721f7a1d03326c009
Commit: c2f65806f62a49f6f3da59e721f7a1d03326c009
Parent: 094488ffab4837fa14069b81a15e0a6bdaca8a73
Author: Alasdair G Kergon <agk@redhat.com>
AuthorDate: Wed Sep 21 16:05:08 2016 +0100
Committer: Alasdair G Kergon <agk@redhat.com>
CommitterDate: Wed Sep 21 16:05:08 2016 +0100
Revert "lvconvert: fix (automatic) raid repair regression"
This reverts commit 5d455b28fce16fce20f33279fe3b1b682c57e619.
This change depends on the kernel version, so should detect
this and behave accordingly.
---
WHATS_NEW | 1 -
daemons/dmeventd/plugins/raid/dmeventd_raid.c | 42 ++--------
lib/metadata/lv.c | 6 ++
lib/metadata/raid_manip.c | 2 +-
test/shell/lvconvert-repair-raid.sh | 104 +-----------------------
tools/lvconvert.c | 18 ++++
6 files changed, 39 insertions(+), 134 deletions(-)
diff --git a/WHATS_NEW b/WHATS_NEW
index f7b39b9..faddb9f 100644
--- a/WHATS_NEW
+++ b/WHATS_NEW
@@ -1,6 +1,5 @@
Version 2.02.166 -
=====================================
- Fix lvconvert --repair regression
Fix reported origin lv field for cache volumes. (2.02.133)
Always specify snapshot cow LV for monitoring not internal LV. (2.02.165)
Fix lvchange --discard|--zero for active thin-pool.
diff --git a/daemons/dmeventd/plugins/raid/dmeventd_raid.c b/daemons/dmeventd/plugins/raid/dmeventd_raid.c
index bec594a..770fbc6 100644
--- a/daemons/dmeventd/plugins/raid/dmeventd_raid.c
+++ b/daemons/dmeventd/plugins/raid/dmeventd_raid.c
@@ -1,5 +1,5 @@
/*
- * Copyright (C) 2005-2016 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2005-2015 Red Hat, Inc. All rights reserved.
*
* This file is part of LVM2.
*
@@ -13,20 +13,14 @@
*/
#include "lib.h"
-#include "defaults.h"
#include "dmeventd_lvm.h"
#include "libdevmapper-event.h"
-/* Hold enough elements for the mximum number of RAID images */
-#define RAID_DEVS_ELEMS ((DEFAULT_RAID_MAX_IMAGES + 63) / 64)
-
struct dso_state {
struct dm_pool *mem;
char cmd_lvscan[512];
char cmd_lvconvert[512];
- uint64_t raid_devs[RAID_DEVS_ELEMS];
int failed;
- int warned;
};
DM_EVENT_LOG_FN("raid")
@@ -37,39 +31,20 @@ static int _process_raid_event(struct dso_state *state, char *params, const char
{
struct dm_status_raid *status;
const char *d;
- int dead = 0, r = 1;
if (!dm_get_status_raid(state->mem, params, &status)) {
log_error("Failed to process status line for %s.", device);
return 0;
}
- d = status->dev_health;
- while ((d = strchr(d, 'D'))) {
- uint32_t dev = (uint32_t)(d - status->dev_health);
-
- if (!(state->raid_devs[dev / 64] & (1 << (dev % 64))))
- log_error("Device #%u of %s array, %s, has failed.",
- dev, status->raid_type, device);
-
- state->raid_devs[dev / 64] |= (1 << (dev % 64));
- d++;
- dead = 1;
- }
-
- if (dead) {
- if (status->insync_regions < status->total_regions) {
- if (!state->warned)
- log_warn("WARNING: waiting for resynchronization to finish "
- "before initiating repair on RAID device %s", device);
-
- state->warned = 1;
- goto out; /* Not yet done syncing with accessible devices */
- }
-
+ if ((d = strchr(status->dev_health, 'D'))) {
if (state->failed)
goto out; /* already reported */
+ log_error("Device #%d of %s array, %s, has failed.",
+ (int)(d - status->dev_health),
+ status->raid_type, device);
+
state->failed = 1;
if (!dmeventd_lvm2_run_with_lock(state->cmd_lvscan))
log_warn("WARNING: Re-scan of RAID device %s failed.", device);
@@ -77,7 +52,8 @@ static int _process_raid_event(struct dso_state *state, char *params, const char
/* if repair goes OK, report success even if lvscan has failed */
if (!dmeventd_lvm2_run_with_lock(state->cmd_lvconvert)) {
log_info("Repair of RAID device %s failed.", device);
- r = 0;
+ dm_pool_free(state->mem, status);
+ return 0;
}
} else {
state->failed = 0;
@@ -88,7 +64,7 @@ static int _process_raid_event(struct dso_state *state, char *params, const char
out:
dm_pool_free(state->mem, status);
- return r;
+ return 1;
}
void process_event(struct dm_task *dmt,
diff --git a/lib/metadata/lv.c b/lib/metadata/lv.c
index 70036f9..03d52ae 100644
--- a/lib/metadata/lv.c
+++ b/lib/metadata/lv.c
@@ -1017,6 +1017,12 @@ int lv_raid_image_in_sync(const struct logical_volume *lv)
return 0;
}
+ if (!lv_raid_percent(raid_seg->lv, &percent))
+ return_0;
+
+ if (percent == DM_PERCENT_100)
+ return 1;
+
/* Find out which sub-LV this is. */
for (s = 0; s < raid_seg->area_count; s++)
if (seg_lv(raid_seg, s) == lv)
diff --git a/lib/metadata/raid_manip.c b/lib/metadata/raid_manip.c
index deb88a2..e5fdf4f 100644
--- a/lib/metadata/raid_manip.c
+++ b/lib/metadata/raid_manip.c
@@ -3658,7 +3658,7 @@ static int _lv_raid_rebuild_or_replace(struct logical_volume *lv,
return 0;
}
- if (!_raid_in_sync(lv)) {
+ if (!mirror_in_sync() && !_raid_in_sync(lv)) {
log_error("Unable to replace devices in %s/%s while it is"
" not in-sync.", lv->vg->name, lv->name);
return 0;
diff --git a/test/shell/lvconvert-repair-raid.sh b/test/shell/lvconvert-repair-raid.sh
index b51d8fe..1ef91c4 100644
--- a/test/shell/lvconvert-repair-raid.sh
+++ b/test/shell/lvconvert-repair-raid.sh
@@ -22,52 +22,11 @@ aux lvmconf 'allocation/maximise_cling = 0' \
aux prepare_vg 8
-function delay
-{
- for d in $(< DEVICES)
- do
- aux delay_dev "$d" 0 $1 $(get first_extent_sector "$d")
- done
-}
-
# It's possible small raid arrays do have problems with reporting in-sync.
# So try bigger size
-RAID_SIZE=32
-
-# Fast sync and repair afterwards
-delay 0
-
-# RAID1 dual-leg single replace after initial sync
-lvcreate --type raid1 -m 1 -L $RAID_SIZE -n $lv1 $vg "$dev1" "$dev2"
-aux wait_for_sync $vg $lv1
-aux disable_dev "$dev2"
-lvconvert -y --repair $vg/$lv1
-vgreduce --removemissing $vg
-aux enable_dev "$dev2"
-vgextend $vg "$dev2"
-lvremove -ff $vg/$lv1
-
-# Delayed sync to allow for repair during rebuild
-delay 50
-
-# RAID1 triple-leg single replace during initial sync
-lvcreate --type raid1 -m 2 -L $RAID_SIZE -n $lv1 $vg "$dev1" "$dev2" "$dev3"
-aux disable_dev "$dev2" "$dev3"
-not lvconvert -y --repair $vg/$lv1
-aux wait_for_sync $vg $lv1
-lvconvert -y --repair $vg/$lv1
-vgreduce --removemissing $vg
-aux enable_dev "$dev2" "$dev3"
-vgextend $vg "$dev2" "$dev3"
-lvremove -ff $vg/$lv1
-
-
-# Larger RAID size possible for striped RAID
RAID_SIZE=64
-# Fast sync and repair afterwards
-delay 0
-# RAID5 single replace after initial sync
+# RAID5 single replace
lvcreate --type raid5 -i 2 -L $RAID_SIZE -n $lv1 $vg "$dev1" "$dev2" "$dev3"
aux wait_for_sync $vg $lv1
aux disable_dev "$dev3"
@@ -75,69 +34,16 @@ lvconvert -y --repair $vg/$lv1
vgreduce --removemissing $vg
aux enable_dev "$dev3"
vgextend $vg "$dev3"
-lvremove -ff $vg/$lv1
+lvremove -ff $vg
-# Delayed sync to allow for repair during rebuild
-delay 50
-
-# RAID5 single replace during initial sync
-lvcreate --type raid5 -i 2 -L $RAID_SIZE -n $lv1 $vg "$dev1" "$dev2" "$dev3"
-aux disable_dev "$dev3"
-not lvconvert -y --repair $vg/$lv1
-aux wait_for_sync $vg $lv1
-lvconvert -y --repair $vg/$lv1
-vgreduce --removemissing $vg
-aux enable_dev "$dev3"
-vgextend $vg "$dev3"
-lvremove -ff $vg/$lv1
-
-# Fast sync and repair afterwards
-delay 0
-
-# RAID6 double replace after initial sync
+# RAID6 double replace
lvcreate --type raid6 -i 3 -L $RAID_SIZE -n $lv1 $vg \
"$dev1" "$dev2" "$dev3" "$dev4" "$dev5"
aux wait_for_sync $vg $lv1
aux disable_dev "$dev4" "$dev5"
lvconvert -y --repair $vg/$lv1
vgreduce --removemissing $vg
-aux enable_dev "$dev4" "$dev5"
-vgextend $vg "$dev4" "$dev5"
-lvremove -ff $vg/$lv1
-
-# Delayed sync to allow for repair during rebuild
-delay 50
-
-# RAID6 single replace after initial sync
-lvcreate --type raid6 -i 3 -L $RAID_SIZE -n $lv1 $vg \
- "$dev1" "$dev2" "$dev3" "$dev4" "$dev5"
-aux disable_dev "$dev4"
-not lvconvert -y --repair $vg/$lv1
-delay 0 # Fast sync and repair afterwards
-aux disable_dev "$dev4" # Need to disable again after changing delay
-aux wait_for_sync $vg $lv1
-lvconvert -y --repair $vg/$lv1
-vgreduce --removemissing $vg
aux enable_dev "$dev4"
-vgextend $vg "$dev4"
-lvremove -ff $vg/$lv1
-
-# Delayed sync to allow for repair during rebuild
-delay 50
-
-# RAID10 single replace after initial sync
-lvcreate --type raid10 -m 1 -i 2 -L $RAID_SIZE -n $lv1 $vg \
- "$dev1" "$dev2" "$dev3" "$dev4"
-aux disable_dev "$dev4"
-not lvconvert -y --repair $vg/$lv1
-delay 0 # Fast sync and repair afterwards
-aux disable_dev "$dev4" # Need to disable again after changing delay
-aux disable_dev "$dev1"
-aux wait_for_sync $vg $lv1
-lvconvert -y --repair $vg/$lv1
-vgreduce --removemissing $vg
-aux enable_dev "$dev4"
-vgextend $vg "$dev4"
-lvremove -ff $vg/$lv1
-
+aux enable_dev "$dev5"
+vgextend $vg "$dev4" "$dev5"
vgremove -ff $vg
diff --git a/tools/lvconvert.c b/tools/lvconvert.c
index 3607247..18010d5 100644
--- a/tools/lvconvert.c
+++ b/tools/lvconvert.c
@@ -1973,6 +1973,24 @@ static int _lvconvert_raid(struct logical_volume *lv, struct lvconvert_params *l
return 0;
}
+ if (!lv_raid_percent(lv, &sync_percent)) {
+ log_error("Unable to determine sync status of %s.",
+ display_lvname(lv));
+ return 0;
+ }
+
+ if (sync_percent != DM_PERCENT_100) {
+ log_warn("WARNING: %s is not in-sync.", display_lvname(lv));
+ log_warn("WARNING: Portions of the array may be unrecoverable.");
+
+ /*
+ * The kernel will not allow a device to be replaced
+ * in an array that is not in-sync unless we override
+ * by forcing the array to be considered "in-sync".
+ */
+ init_mirror_in_sync(1);
+ }
+
_lvconvert_raid_repair_ask(cmd, lp, &replace);
if (replace) {
^ permalink raw reply related [flat|nested] only message in thread
only message in thread, other threads:[~2016-09-21 15:06 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-09-21 15:06 master - Revert "lvconvert: fix (automatic) raid repair regression" Alasdair Kergon
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.