|
 |
21ab4e |
From 8c6bf26a112521284a62082f5796df8f9b504f54 Mon Sep 17 00:00:00 2001
|
|
|
21ab4e |
From: karthik-us <ksubrahm@redhat.com>
|
|
|
21ab4e |
Date: Fri, 31 Mar 2017 15:17:42 +0530
|
|
|
21ab4e |
Subject: [PATCH 369/369] cluster/afr: Undo pending xattrs only on the up
|
|
|
21ab4e |
bricks
|
|
|
21ab4e |
|
|
|
21ab4e |
Problem:
|
|
|
21ab4e |
While doing conservative merge, even if a brick is down, it will reset
|
|
|
21ab4e |
the pending xattr on that. When that brick comes up, as part of the
|
|
|
21ab4e |
heal, it will consider this brick as the source and removes the entries
|
|
|
21ab4e |
on the other bricks, which leads to data loss.
|
|
|
21ab4e |
|
|
|
21ab4e |
Fix: Undo pending only for the bricks which are up.
|
|
|
21ab4e |
|
|
|
21ab4e |
> Reviewed-on: https://review.gluster.org/16913
|
|
|
21ab4e |
> Reviewed-by: Pranith Kumar Karampuri <pkarampu@redhat.com>
|
|
|
21ab4e |
> Smoke: Gluster Build System <jenkins@build.gluster.org>
|
|
|
21ab4e |
> NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
|
|
|
21ab4e |
> CentOS-regression: Gluster Build System <jenkins@build.gluster.org>
|
|
|
21ab4e |
> Reviewed-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
21ab4e |
|
|
|
21ab4e |
Change-Id: Ie166ac2e738f0b231b42082855f3c1c57797e7a7
|
|
|
21ab4e |
BUG: 1437773
|
|
|
21ab4e |
Signed-off-by: karthik-us <ksubrahm@redhat.com>
|
|
|
21ab4e |
Reviewed-on: https://code.engineering.redhat.com/gerrit/102101
|
|
|
21ab4e |
Reviewed-by: Ravishankar Narayanankutty <ravishankar@redhat.com>
|
|
|
21ab4e |
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>
|
|
|
21ab4e |
---
|
|
|
21ab4e |
.../bug-1433571-undo-pending-only-on-up-bricks.t | 89 ++++++++++++++++++++++
|
|
|
21ab4e |
xlators/cluster/afr/src/afr-self-heal-common.c | 2 +-
|
|
|
21ab4e |
2 files changed, 90 insertions(+), 1 deletion(-)
|
|
|
21ab4e |
create mode 100644 tests/bugs/replicate/bug-1433571-undo-pending-only-on-up-bricks.t
|
|
|
21ab4e |
|
|
|
21ab4e |
diff --git a/tests/bugs/replicate/bug-1433571-undo-pending-only-on-up-bricks.t b/tests/bugs/replicate/bug-1433571-undo-pending-only-on-up-bricks.t
|
|
|
21ab4e |
new file mode 100644
|
|
|
21ab4e |
index 0000000..271abb4
|
|
|
21ab4e |
--- /dev/null
|
|
|
21ab4e |
+++ b/tests/bugs/replicate/bug-1433571-undo-pending-only-on-up-bricks.t
|
|
|
21ab4e |
@@ -0,0 +1,89 @@
|
|
|
21ab4e |
+#!/bin/bash
|
|
|
21ab4e |
+. $(dirname $0)/../../include.rc
|
|
|
21ab4e |
+. $(dirname $0)/../../volume.rc
|
|
|
21ab4e |
+cleanup;
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+TEST glusterd
|
|
|
21ab4e |
+TEST pidof glusterd
|
|
|
21ab4e |
+TEST $CLI volume create $V0 replica 3 $H0:$B0/${V0}{0,1,2}
|
|
|
21ab4e |
+TEST $CLI volume start $V0
|
|
|
21ab4e |
+TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0;
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+# Disable self-heal-daemon, client-side-heal and set quorum-type to none
|
|
|
21ab4e |
+TEST $CLI volume set $V0 cluster.self-heal-daemon off
|
|
|
21ab4e |
+TEST $CLI volume set $V0 cluster.data-self-heal off
|
|
|
21ab4e |
+TEST $CLI volume set $V0 cluster.metadata-self-heal off
|
|
|
21ab4e |
+TEST $CLI volume set $V0 cluster.entry-self-heal off
|
|
|
21ab4e |
+TEST $CLI volume set $V0 cluster.quorum-type none
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+#Kill bricks 0 & 1 and create a file to have pending entry for 0 & 1 on brick 2
|
|
|
21ab4e |
+TEST kill_brick $V0 $H0 $B0/${V0}0
|
|
|
21ab4e |
+TEST kill_brick $V0 $H0 $B0/${V0}1
|
|
|
21ab4e |
+echo "file 1" >> $M0/f1
|
|
|
21ab4e |
+EXPECT "000000000000000000000001" get_hex_xattr trusted.afr.$V0-client-0 $B0/${V0}2
|
|
|
21ab4e |
+EXPECT "000000000000000000000001" get_hex_xattr trusted.afr.$V0-client-1 $B0/${V0}2
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+TEST $CLI volume start $V0 force
|
|
|
21ab4e |
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status $V0 0
|
|
|
21ab4e |
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status $V0 1
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+#Kill bricks 1 & 2 and create a file to have pending entry for 1 & 2 on brick 0
|
|
|
21ab4e |
+TEST kill_brick $V0 $H0 $B0/${V0}1
|
|
|
21ab4e |
+TEST kill_brick $V0 $H0 $B0/${V0}2
|
|
|
21ab4e |
+echo "file 2" >> $M0/f2
|
|
|
21ab4e |
+EXPECT "000000000000000000000001" get_hex_xattr trusted.afr.$V0-client-1 $B0/${V0}0
|
|
|
21ab4e |
+EXPECT "000000000000000000000001" get_hex_xattr trusted.afr.$V0-client-2 $B0/${V0}0
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+TEST $CLI volume start $V0 force
|
|
|
21ab4e |
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status $V0 1
|
|
|
21ab4e |
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status $V0 2
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+#Kill bricks 2 & 0 and create a file to have pending entry for 2 & 0 on brick 1
|
|
|
21ab4e |
+TEST kill_brick $V0 $H0 $B0/${V0}2
|
|
|
21ab4e |
+TEST kill_brick $V0 $H0 $B0/${V0}0
|
|
|
21ab4e |
+echo "file 3" >> $M0/f3
|
|
|
21ab4e |
+EXPECT "000000000000000000000001" get_hex_xattr trusted.afr.$V0-client-0 $B0/${V0}1
|
|
|
21ab4e |
+EXPECT "000000000000000000000001" get_hex_xattr trusted.afr.$V0-client-2 $B0/${V0}1
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+TEST $CLI volume start $V0 force
|
|
|
21ab4e |
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status $V0 0
|
|
|
21ab4e |
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status $V0 2
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+#Kill brick 0 and turn on the client side heal and do ls to trigger the heal.
|
|
|
21ab4e |
+#The pending xattrs on bricks 1 & 2 should have pending entry on brick 0.
|
|
|
21ab4e |
+TEST kill_brick $V0 $H0 $B0/${V0}0
|
|
|
21ab4e |
+TEST $CLI volume set $V0 cluster.data-self-heal on
|
|
|
21ab4e |
+TEST $CLI volume set $V0 cluster.metadata-self-heal on
|
|
|
21ab4e |
+TEST $CLI volume set $V0 cluster.entry-self-heal on
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+TEST ls $M0
|
|
|
21ab4e |
+EXPECT "000000000000000000000001" get_hex_xattr trusted.afr.$V0-client-0 $B0/${V0}1
|
|
|
21ab4e |
+EXPECT "000000000000000000000000" get_hex_xattr trusted.afr.$V0-client-2 $B0/${V0}1
|
|
|
21ab4e |
+EXPECT "000000000000000000000001" get_hex_xattr trusted.afr.$V0-client-0 $B0/${V0}2
|
|
|
21ab4e |
+EXPECT "000000000000000000000000" get_hex_xattr trusted.afr.$V0-client-1 $B0/${V0}2
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+#Bring back all the bricks and trigger the heal again by doing ls. Now the
|
|
|
21ab4e |
+#pending xattrs on all the bricks should be 0.
|
|
|
21ab4e |
+TEST $CLI volume start $V0 force
|
|
|
21ab4e |
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status $V0 0
|
|
|
21ab4e |
+TEST ls $M0
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+TEST cat $M0/f1
|
|
|
21ab4e |
+TEST cat $M0/f2
|
|
|
21ab4e |
+TEST cat $M0/f3
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+EXPECT_WITHIN $HEAL_TIMEOUT "^0$" get_pending_heal_count $V0
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+EXPECT "000000000000000000000000" get_hex_xattr trusted.afr.$V0-client-1 $B0/${V0}0
|
|
|
21ab4e |
+EXPECT "000000000000000000000000" get_hex_xattr trusted.afr.$V0-client-2 $B0/${V0}0
|
|
|
21ab4e |
+EXPECT "000000000000000000000000" get_hex_xattr trusted.afr.$V0-client-0 $B0/${V0}1
|
|
|
21ab4e |
+EXPECT "000000000000000000000000" get_hex_xattr trusted.afr.$V0-client-2 $B0/${V0}1
|
|
|
21ab4e |
+EXPECT "000000000000000000000000" get_hex_xattr trusted.afr.$V0-client-0 $B0/${V0}2
|
|
|
21ab4e |
+EXPECT "000000000000000000000000" get_hex_xattr trusted.afr.$V0-client-1 $B0/${V0}2
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+#Check whether all the bricks contains all the 3 files.
|
|
|
21ab4e |
+EXPECT "3" echo $(ls $B0/${V0}0 | wc -l)
|
|
|
21ab4e |
+EXPECT "3" echo $(ls $B0/${V0}1 | wc -l)
|
|
|
21ab4e |
+EXPECT "3" echo $(ls $B0/${V0}2 | wc -l)
|
|
|
21ab4e |
+
|
|
|
21ab4e |
+cleanup;
|
|
|
21ab4e |
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c
|
|
|
21ab4e |
index 8b2c3fa..9f6cbcd 100644
|
|
|
21ab4e |
--- a/xlators/cluster/afr/src/afr-self-heal-common.c
|
|
|
21ab4e |
+++ b/xlators/cluster/afr/src/afr-self-heal-common.c
|
|
|
21ab4e |
@@ -205,7 +205,7 @@ afr_selfheal_undo_pending (call_frame_t *frame, xlator_t *this, inode_t *inode,
|
|
|
21ab4e |
output_matrix[i][j] = 1;
|
|
|
21ab4e |
if (type == AFR_ENTRY_TRANSACTION)
|
|
|
21ab4e |
full_heal_mtx_out[i][j] = 1;
|
|
|
21ab4e |
- } else {
|
|
|
21ab4e |
+ } else if (locked_on[j]) {
|
|
|
21ab4e |
output_matrix[i][j] = -input_matrix[i][j];
|
|
|
21ab4e |
if (type == AFR_ENTRY_TRANSACTION)
|
|
|
21ab4e |
full_heal_mtx_out[i][j] = -full_heal_mtx_in[i][j];
|
|
|
21ab4e |
--
|
|
|
21ab4e |
1.8.3.1
|
|
|
21ab4e |
|