summaryrefslogtreecommitdiffstats
path: root/xlators
diff options
context:
space:
mode:
authorRavishankar N <ravishankar@redhat.com>2020-02-11 14:34:48 +0530
committerRavishankar N <ravishankar@redhat.com>2020-02-18 09:11:40 +0000
commit06453d77d056fbaa393a137ca277a20e38d2f67e (patch)
tree596d5803946caaebc018758508593ba2232a2ce1 /xlators
parenta7fa54ddea3fe429f143b37e4de06a93b49d776a (diff)
downloadglusterfs-06453d77d056fbaa393a137ca277a20e38d2f67e.tar.gz
glusterfs-06453d77d056fbaa393a137ca277a20e38d2f67e.tar.xz
glusterfs-06453d77d056fbaa393a137ca277a20e38d2f67e.zip
afr: prevent spurious entry heals leading to gfid split-brain
Problem: In a hyperconverged setup with granular-entry-heal enabled, if a file is recreated while one of the bricks is down, and an index heal is triggered (with the brick still down), entry-self heal was doing a spurious heal with just the 2 good bricks. It was doing a post-op leading to removal of the filename from .glusterfs/indices/entry-changes as well as erroneous setting of afr xattrs on the parent. When the brick came up, the xattrs were cleared, resulting in the renamed file not getting healed and leading to gfid split-brain and EIO on the mount. Fix: Proceed with entry heal only when shd can connect to all bricks of the replica, just like in data and metadata heal. fixes: bz#1801624 Change-Id: I916ae26ad1fabf259bc6362da52d433b7223b17e Signed-off-by: Ravishankar N <ravishankar@redhat.com>
Diffstat (limited to 'xlators')
-rw-r--r--xlators/cluster/afr/src/afr-common.c4
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-common.c8
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-entry.c6
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-name.c2
-rw-r--r--xlators/cluster/afr/src/afr-self-heal.h2
5 files changed, 7 insertions, 15 deletions
diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index 83246a2418..5462f0c7f5 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -7423,7 +7423,7 @@ afr_fav_child_reset_sink_xattrs(void *opaque)
ret = afr_selfheal_inodelk(heal_frame, this, inode, this->name, 0, 0,
locked_on);
{
- if (ret < AFR_SH_MIN_PARTICIPANTS)
+ if (ret < priv->child_count)
goto data_unlock;
ret = __afr_selfheal_data_prepare(
heal_frame, this, inode, locked_on, sources, sinks,
@@ -7440,7 +7440,7 @@ afr_fav_child_reset_sink_xattrs(void *opaque)
ret = afr_selfheal_inodelk(heal_frame, this, inode, this->name,
LLONG_MAX - 1, 0, locked_on);
{
- if (ret < AFR_SH_MIN_PARTICIPANTS)
+ if (ret < priv->child_count)
goto mdata_unlock;
ret = __afr_selfheal_metadata_prepare(
heal_frame, this, inode, locked_on, sources, sinks,
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c
index 16f19e7bc6..b2f0ef4801 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.c
+++ b/xlators/cluster/afr/src/afr-self-heal-common.c
@@ -1575,7 +1575,6 @@ afr_selfheal_find_direction(call_frame_t *frame, xlator_t *this,
char *accused = NULL; /* Accused others without any self-accusal */
char *pending = NULL; /* Have pending operations on others */
char *self_accused = NULL; /* Accused itself */
- int min_participants = -1;
priv = this->private;
@@ -1599,12 +1598,7 @@ afr_selfheal_find_direction(call_frame_t *frame, xlator_t *this,
}
}
- if (type == AFR_DATA_TRANSACTION || type == AFR_METADATA_TRANSACTION) {
- min_participants = priv->child_count;
- } else {
- min_participants = AFR_SH_MIN_PARTICIPANTS;
- }
- if (afr_success_count(replies, priv->child_count) < min_participants) {
+ if (afr_success_count(replies, priv->child_count) < priv->child_count) {
/* Treat this just like locks not being acquired */
return -ENOTCONN;
}
diff --git a/xlators/cluster/afr/src/afr-self-heal-entry.c b/xlators/cluster/afr/src/afr-self-heal-entry.c
index e7062289c7..ac31751997 100644
--- a/xlators/cluster/afr/src/afr-self-heal-entry.c
+++ b/xlators/cluster/afr/src/afr-self-heal-entry.c
@@ -598,7 +598,7 @@ afr_selfheal_entry_dirent(call_frame_t *frame, xlator_t *this, fd_t *fd,
ret = afr_selfheal_entrylk(frame, this, fd->inode, this->name, NULL,
locked_on);
{
- if (ret < AFR_SH_MIN_PARTICIPANTS) {
+ if (ret < priv->child_count) {
gf_msg_debug(this->name, 0,
"%s: Skipping "
"entry self-heal as only %d sub-volumes "
@@ -992,7 +992,7 @@ __afr_selfheal_entry(call_frame_t *frame, xlator_t *this, fd_t *fd,
ret = afr_selfheal_entrylk(frame, this, fd->inode, this->name, NULL,
data_lock);
{
- if (ret < AFR_SH_MIN_PARTICIPANTS) {
+ if (ret < priv->child_count) {
gf_msg_debug(this->name, 0,
"%s: Skipping "
"entry self-heal as only %d sub-volumes could "
@@ -1116,7 +1116,7 @@ afr_selfheal_entry(call_frame_t *frame, xlator_t *this, inode_t *inode)
ret = afr_selfheal_tie_breaker_entrylk(frame, this, inode, priv->sh_domain,
NULL, locked_on);
{
- if (ret < AFR_SH_MIN_PARTICIPANTS) {
+ if (ret < priv->child_count) {
gf_msg_debug(this->name, 0,
"%s: Skipping "
"entry self-heal as only %d sub-volumes could "
diff --git a/xlators/cluster/afr/src/afr-self-heal-name.c b/xlators/cluster/afr/src/afr-self-heal-name.c
index 36640b5456..7d4f2080ec 100644
--- a/xlators/cluster/afr/src/afr-self-heal-name.c
+++ b/xlators/cluster/afr/src/afr-self-heal-name.c
@@ -514,7 +514,7 @@ afr_selfheal_name_do(call_frame_t *frame, xlator_t *this, inode_t *parent,
ret = afr_selfheal_entrylk(frame, this, parent, this->name, bname,
locked_on);
{
- if (ret < AFR_SH_MIN_PARTICIPANTS) {
+ if (ret < priv->child_count) {
ret = -ENOTCONN;
goto unlock;
}
diff --git a/xlators/cluster/afr/src/afr-self-heal.h b/xlators/cluster/afr/src/afr-self-heal.h
index 5e7bde8689..9746860e36 100644
--- a/xlators/cluster/afr/src/afr-self-heal.h
+++ b/xlators/cluster/afr/src/afr-self-heal.h
@@ -11,8 +11,6 @@
#ifndef _AFR_SELFHEAL_H
#define _AFR_SELFHEAL_H
-#define AFR_SH_MIN_PARTICIPANTS 2
-
/* Perform fop on all UP subvolumes and wait for all callbacks to return */
#define AFR_ONALL(frame, rfn, fop, args...) \