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#1804594
Change-Id: I916ae26ad1fabf259bc6362da52d433b7223b17e
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
(cherry picked from commit 06453d77d056fbaa393a137ca277a20e38d2f67e)

1 files changed, 58 insertions, 0 deletions

diff --git a/tests/bugs/replicate/bug-1801624-entry-heal.t b/tests/bugs/replicate/bug-1801624-entry-heal.t
new file mode 100644
index 00000000000..94b465181fa
--- /dev/null
+++ b/tests/bugs/replicate/bug-1801624-entry-heal.t
@@ -0,0 +1,58 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+cleanup;
+
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume create $V0 replica 3 $H0:$B0/brick{0,1,2}
+TEST $CLI volume set $V0 heal-timeout 5
+TEST $CLI volume start $V0
+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 $CLI volume heal $V0 granular-entry-heal enable
+
+TEST glusterfs --volfile-id=/$V0 --volfile-server=$H0 $M0 --attribute-timeout=0 --entry-timeout=0
+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 $V0 2
+echo "Data">$M0/FILE
+ret=$?
+TEST [ $ret -eq 0 ]
+
+# Re-create the file when a brick is down.
+TEST kill_brick $V0 $H0 $B0/brick1
+TEST rm $M0/FILE
+echo "New Data">$M0/FILE
+ret=$?
+TEST [ $ret -eq 0 ]
+EXPECT_WITHIN $HEAL_TIMEOUT "4" get_pending_heal_count $V0
+
+# Launching index heal must not reset parent dir afr xattrs or remove granular entry indices.
+$CLI volume heal $V0 # CLI will fail but heal is launched anyway.
+TEST sleep 5 # give index heal a chance to do one run.
+brick0_pending=$(get_hex_xattr trusted.afr.$V0-client-1 $B0/brick0/)
+brick2_pending=$(get_hex_xattr trusted.afr.$V0-client-1 $B0/brick2/)
+TEST [ $brick0_pending -eq "000000000000000000000002" ]
+TEST [ $brick2_pending -eq "000000000000000000000002" ]
+EXPECT "FILE" ls $B0/brick0/.glusterfs/indices/entry-changes/00000000-0000-0000-0000-000000000001/
+EXPECT "FILE" ls $B0/brick2/.glusterfs/indices/entry-changes/00000000-0000-0000-0000-000000000001/
+
+TEST $CLI volume start $V0 force
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" brick_up_status $V0 $H0 $B0/brick1
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status_in_shd $V0 1
+$CLI volume heal $V0
+EXPECT_WITHIN $HEAL_TIMEOUT "0" get_pending_heal_count $V0
+
+# No gfid-split-brain (i.e. EIO) must be seen. Try on fresh mount to avoid cached values.
+EXPECT_WITHIN $UMOUNT_TIMEOUT "Y" force_umount $M0
+TEST glusterfs --volfile-id=/$V0 --volfile-server=$H0 $M0 --attribute-timeout=0 --entry-timeout=0
+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
+TEST cat $M0/FILE
+
+EXPECT_WITHIN $UMOUNT_TIMEOUT "Y" force_umount $M0
+cleanup;