From 6fb1804d113ae996e085ef0f23fa8908d167f006 Mon Sep 17 00:00:00 2001
From: Ravishankar N <ravishankar@redhat.com>
Date: Thu, 14 Jun 2018 12:59:06 +0530
Subject: [PATCH 299/305] afr: heal gfids when file is not present on all
bricks
Backport of: https://review.gluster.org/#/c/20271/
commit f73814ad08d552d94d0139b2592175d206e7a166 (rhgs) introduced a regression
wherein if a file is present in only 1 brick of replica *and* doesn't
have a gfid associated with it, it doesn't get healed upon the next
lookup from the client. Fix it.
Change-Id: I7d1111dcb45b1b8b8340a7d02558f05df70aa599
BUG: 1592666
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/141899
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Karthik Subrahmanya <ksubrahm@redhat.com>
---
.../replicate/bug-1591193-assign-gfid-and-heal.t | 128 +++++++++++++++++++++
xlators/cluster/afr/src/afr-self-heal-common.c | 39 ++++++-
xlators/cluster/afr/src/afr-self-heal-data.c | 8 +-
xlators/cluster/afr/src/afr-self-heal-entry.c | 4 +-
xlators/cluster/afr/src/afr-self-heal-name.c | 6 +-
xlators/cluster/afr/src/afr-self-heal.h | 6 +-
6 files changed, 179 insertions(+), 12 deletions(-)
create mode 100644 tests/bugs/replicate/bug-1591193-assign-gfid-and-heal.t
diff --git a/tests/bugs/replicate/bug-1591193-assign-gfid-and-heal.t b/tests/bugs/replicate/bug-1591193-assign-gfid-and-heal.t
new file mode 100644
index 0000000..d3b5f9a
--- /dev/null
+++ b/tests/bugs/replicate/bug-1591193-assign-gfid-and-heal.t
@@ -0,0 +1,128 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+. $(dirname $0)/../../afr.rc
+
+cleanup;
+
+function check_gfid_and_link_count
+{
+ local file=$1
+
+ file_gfid_b0=$(gf_get_gfid_xattr $B0/${V0}0/$file)
+ TEST [ ! -z $file_gfid_b0 ]
+ file_gfid_b1=$(gf_get_gfid_xattr $B0/${V0}1/$file)
+ file_gfid_b2=$(gf_get_gfid_xattr $B0/${V0}2/$file)
+ EXPECT $file_gfid_b0 echo $file_gfid_b1
+ EXPECT $file_gfid_b0 echo $file_gfid_b2
+
+ EXPECT "2" stat -c %h $B0/${V0}0/$file
+ EXPECT "2" stat -c %h $B0/${V0}1/$file
+ EXPECT "2" stat -c %h $B0/${V0}2/$file
+}
+TESTS_EXPECTED_IN_LOOP=30
+
+##############################################################################
+# Test on 1x3 volume
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume create $V0 replica 3 $H0:$B0/${V0}{0,1,2};
+TEST $CLI volume start $V0;
+
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" brick_up_status $V0 $H0 $B0/${V0}0
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" brick_up_status $V0 $H0 $B0/${V0}1
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" brick_up_status $V0 $H0 $B0/${V0}2
+
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "Y" glustershd_up_status
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 0
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 1
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 2
+
+TEST glusterfs --volfile-id=/$V0 --volfile-server=$H0 --attribute-timeout=0 --entry-timeout=0 $M0
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" afr_child_up_status $V0 0
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" afr_child_up_status $V0 1
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" afr_child_up_status $V0 2
+
+
+# Create files directly in the backend on different bricks
+echo $RANDOM >> $B0/${V0}0/file1
+echo $RANDOM >> $B0/${V0}1/file2
+echo $RANDOM >> $B0/${V0}2/file3
+
+# To prevent is_fresh_file code path
+sleep 2
+
+# Access them from mount to trigger name + gfid heal.
+TEST stat $M0/file1
+TEST stat $M0/file2
+TEST stat $M0/file3
+
+# Launch index heal to complete any pending data/metadata heals.
+TEST $CLI volume heal $V0
+EXPECT_WITHIN $HEAL_TIMEOUT "^0$" get_pending_heal_count $V0
+
+# Check each file has a gfid and the .glusterfs hardlink
+check_gfid_and_link_count file1
+check_gfid_and_link_count file2
+check_gfid_and_link_count file3
+
+TEST rm $M0/file1
+TEST rm $M0/file2
+TEST rm $M0/file3
+cleanup;
+
+##############################################################################
+# Test on 1x (2+1) volume
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume create $V0 replica 3 arbiter 1 $H0:$B0/${V0}{0,1,2};
+TEST $CLI volume start $V0;
+
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" brick_up_status $V0 $H0 $B0/${V0}0
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" brick_up_status $V0 $H0 $B0/${V0}1
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" brick_up_status $V0 $H0 $B0/${V0}2
+
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "Y" glustershd_up_status
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 0
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 1
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 2
+
+TEST glusterfs --volfile-id=/$V0 --volfile-server=$H0 --attribute-timeout=0 --entry-timeout=0 $M0
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" afr_child_up_status $V0 0
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" afr_child_up_status $V0 1
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" afr_child_up_status $V0 2
+
+
+# Create files directly in the backend on different bricks
+echo $RANDOM >> $B0/${V0}0/file1
+echo $RANDOM >> $B0/${V0}1/file2
+touch $B0/${V0}2/file3
+
+# To prevent is_fresh_file code path
+sleep 2
+
+# Access them from mount to trigger name + gfid heal.
+TEST stat $M0/file1
+TEST stat $M0/file2
+
+# Though file is created on all 3 bricks, lookup will fail as arbiter blames the
+# other 2 bricks and ariter is not 'readable'.
+# TEST ! stat $M0/file3
+# But the checks for failing lookups when quorum is not met is not yet there in
+# rhgs-3.4.0, so stat will succeed.
+TEST stat $M0/file3
+
+# Launch index heal to complete any pending data/metadata heals.
+TEST $CLI volume heal $V0
+EXPECT_WITHIN $HEAL_TIMEOUT "^0$" get_pending_heal_count $V0
+
+# Check each file has a gfid and the .glusterfs hardlink
+check_gfid_and_link_count file1
+check_gfid_and_link_count file2
+check_gfid_and_link_count file3
+
+TEST rm $M0/file1
+TEST rm $M0/file2
+TEST rm $M0/file3
+cleanup;
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c
index 32fd24a..50989d6 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.c
+++ b/xlators/cluster/afr/src/afr-self-heal-common.c
@@ -22,7 +22,7 @@ afr_heal_synctask (xlator_t *this, afr_local_t *local);
int
afr_lookup_and_heal_gfid (xlator_t *this, inode_t *parent, const char *name,
inode_t *inode, struct afr_reply *replies,
- int source, void *gfid)
+ int source, unsigned char *sources, void *gfid)
{
afr_private_t *priv = NULL;
call_frame_t *frame = NULL;
@@ -37,6 +37,23 @@ afr_lookup_and_heal_gfid (xlator_t *this, inode_t *parent, const char *name,
priv = this->private;
wind_on = alloca0 (priv->child_count);
ia_type = replies[source].poststat.ia_type;
+ if ((ia_type == IA_INVAL) &&
+ (AFR_COUNT(sources, priv->child_count) == priv->child_count)) {
+ /* If a file is present on some bricks of the replica but parent
+ * dir does not have pending xattrs, all bricks are sources and
+ * the 'source' we selected earlier might be one where the file
+ * is not actually present. Hence check if file is present in
+ * any of the sources.*/
+ for (i = 0; i < priv->child_count; i++) {
+ if (i == source)
+ continue;
+ if (sources[i] && replies[i].valid &&
+ replies[i].op_ret == 0) {
+ ia_type = replies[i].poststat.ia_type;
+ break;
+ }
+ }
+ }
/* gfid heal on those subvolumes that do not have gfid associated
* with the inode and update those replies.
@@ -1250,6 +1267,21 @@ afr_mark_split_brain_source_sinks_by_policy (call_frame_t *frame,
return fav_child;
}
+gf_boolean_t
+afr_is_file_empty_on_all_children (afr_private_t *priv,
+ struct afr_reply *replies)
+{
+ int i = 0;
+
+ for (i = 0; i < priv->child_count; i++) {
+ if ((!replies[i].valid) || (replies[i].op_ret != 0) ||
+ (replies[i].poststat.ia_size != 0))
+ return _gf_false;
+ }
+
+ return _gf_true;
+}
+
int
afr_mark_source_sinks_if_file_empty (xlator_t *this, unsigned char *sources,
unsigned char *sinks,
@@ -1268,11 +1300,8 @@ afr_mark_source_sinks_if_file_empty (xlator_t *this, unsigned char *sources,
return -1;
if (type == AFR_DATA_TRANSACTION) {
- for (i = 0; i < priv->child_count; i++) {
- if (replies[i].poststat.ia_size != 0)
+ if (!afr_is_file_empty_on_all_children(priv, replies))
return -1;
- }
-
goto mark;
}
diff --git a/xlators/cluster/afr/src/afr-self-heal-data.c b/xlators/cluster/afr/src/afr-self-heal-data.c
index f872a98..3ef7376 100644
--- a/xlators/cluster/afr/src/afr-self-heal-data.c
+++ b/xlators/cluster/afr/src/afr-self-heal-data.c
@@ -670,6 +670,7 @@ __afr_selfheal_data (call_frame_t *frame, xlator_t *this, fd_t *fd,
int source = -1;
gf_boolean_t did_sh = _gf_true;
gf_boolean_t is_arbiter_the_only_sink = _gf_false;
+ gf_boolean_t empty_file = _gf_false;
priv = this->private;
@@ -710,6 +711,11 @@ __afr_selfheal_data (call_frame_t *frame, xlator_t *this, fd_t *fd,
source = ret;
if (AFR_IS_ARBITER_BRICK(priv, source)) {
+ empty_file = afr_is_file_empty_on_all_children (priv,
+ locked_replies);
+ if (empty_file)
+ goto restore_time;
+
did_sh = _gf_false;
goto unlock;
}
@@ -746,7 +752,7 @@ restore_time:
afr_selfheal_restore_time (frame, this, fd->inode, source,
healed_sinks, locked_replies);
- if (!is_arbiter_the_only_sink) {
+ if (!is_arbiter_the_only_sink || !empty_file) {
ret = afr_selfheal_inodelk (frame, this, fd->inode, this->name,
0, 0, data_lock);
if (ret < priv->child_count) {
diff --git a/xlators/cluster/afr/src/afr-self-heal-entry.c b/xlators/cluster/afr/src/afr-self-heal-entry.c
index 647dd71..f6d3a8a 100644
--- a/xlators/cluster/afr/src/afr-self-heal-entry.c
+++ b/xlators/cluster/afr/src/afr-self-heal-entry.c
@@ -187,7 +187,7 @@ __afr_selfheal_heal_dirent (call_frame_t *frame, xlator_t *this, fd_t *fd,
if (replies[source].op_ret == 0) {
ret = afr_lookup_and_heal_gfid (this, fd->inode, name,
- inode, replies, source,
+ inode, replies, source, sources,
&replies[source].poststat.ia_gfid);
if (ret)
return ret;
@@ -320,7 +320,7 @@ __afr_selfheal_merge_dirent (call_frame_t *frame, xlator_t *this, fd_t *fd,
}
ret = afr_lookup_and_heal_gfid (this, fd->inode, name, inode, replies,
- source,
+ source, sources,
&replies[source].poststat.ia_gfid);
if (ret)
return ret;
diff --git a/xlators/cluster/afr/src/afr-self-heal-name.c b/xlators/cluster/afr/src/afr-self-heal-name.c
index 556d14b..bcd0e60 100644
--- a/xlators/cluster/afr/src/afr-self-heal-name.c
+++ b/xlators/cluster/afr/src/afr-self-heal-name.c
@@ -19,7 +19,7 @@ __afr_selfheal_assign_gfid (xlator_t *this, inode_t *parent, uuid_t pargfid,
const char *bname, inode_t *inode,
struct afr_reply *replies, void *gfid,
unsigned char *locked_on, int source,
- gf_boolean_t is_gfid_absent)
+ unsigned char *sources, gf_boolean_t is_gfid_absent)
{
int ret = 0;
int up_count = 0;
@@ -48,7 +48,7 @@ __afr_selfheal_assign_gfid (xlator_t *this, inode_t *parent, uuid_t pargfid,
}
afr_lookup_and_heal_gfid (this, parent, bname, inode, replies, source,
- gfid);
+ sources, gfid);
out:
return ret;
@@ -426,7 +426,7 @@ __afr_selfheal_name_do (call_frame_t *frame, xlator_t *this, inode_t *parent,
is_gfid_absent = (gfid_idx == -1) ? _gf_true : _gf_false;
ret = __afr_selfheal_assign_gfid (this, parent, pargfid, bname, inode,
replies, gfid, locked_on, source,
- is_gfid_absent);
+ sources, is_gfid_absent);
if (ret)
return ret;
diff --git a/xlators/cluster/afr/src/afr-self-heal.h b/xlators/cluster/afr/src/afr-self-heal.h
index b015976..cc99d9e 100644
--- a/xlators/cluster/afr/src/afr-self-heal.h
+++ b/xlators/cluster/afr/src/afr-self-heal.h
@@ -113,7 +113,7 @@ afr_selfheal_entry (call_frame_t *frame, xlator_t *this, inode_t *inode);
int
afr_lookup_and_heal_gfid (xlator_t *this, inode_t *parent, const char *name,
inode_t *inode, struct afr_reply *replies, int source,
- void *gfid);
+ unsigned char *sources, void *gfid);
int
afr_selfheal_inodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
@@ -354,4 +354,8 @@ afr_mark_source_sinks_if_file_empty (xlator_t *this, unsigned char *sources,
struct afr_reply *replies,
afr_transaction_type type);
+gf_boolean_t
+afr_is_file_empty_on_all_children (afr_private_t *priv,
+ struct afr_reply *replies);
+
#endif /* !_AFR_SELFHEAL_H */
--
1.8.3.1