glusterd: Make reset-brick work correctly if brick-mux is on

Reset brick currently kills of the corresponding brick process.
However, with brick multiplexing enabled, stopping the brick
process would render all bricks attached to it unavailable. To
handle this correctly, we need to make sure that the brick process
is terminated only if brick-multiplexing is disabled. Otherwise,
we should send the GLUSTERD_BRICK_TERMINATE rpc to the respective
brick process to detach the brick that is to be reset.

Change-Id: I69002d66ffe6ec36ef48af09b66c522c6d35ac58
BUG: 1446172
Signed-off-by: Samikshan Bairagya <samikshan@gmail.com>
Reviewed-on: https://review.gluster.org/17128
Smoke: Gluster Build System <jenkins@build.gluster.org>
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
CentOS-regression: Gluster Build System <jenkins@build.gluster.org>
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>

2 files changed, 85 insertions, 1 deletions

diff --git a/tests/bugs/glusterd/bug-1446172-brick-mux-reset-brick.t b/tests/bugs/glusterd/bug-1446172-brick-mux-reset-brick.t
new file mode 100644
index 00000000000..e6aaaa4e87c
--- /dev/null
+++ b/tests/bugs/glusterd/bug-1446172-brick-mux-reset-brick.t
@@ -0,0 +1,79 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../traps.rc
+. $(dirname $0)/../../volume.rc
+
+cleanup;
+
+function count_up_bricks {
+        $CLI --xml volume status | grep '<status>1' | wc -l
+}
+
+function count_brick_processes {
+	pgrep glusterfsd | wc -l
+}
+
+TEST glusterd
+
+TEST $CLI volume set all cluster.brick-multiplex on
+push_trapfunc "$CLI volume set all cluster.brick-multiplex off"
+push_trapfunc "cleanup"
+
+TEST $CLI volume create $V0 $H0:$B0/${V0}{0,1}
+TEST $CLI volume create $V1 $H0:$B0/${V1}{0,1}
+
+TEST $CLI volume start $V0
+TEST $CLI volume start $V1
+
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 4 count_up_bricks
+EXPECT 1 count_brick_processes
+
+TEST glusterfs --volfile-id=$V0 --volfile-server=$H0 $M0;
+# Create files
+for i in {1..5}
+do
+        echo $i > $M0/file$i.txt
+done
+
+TEST $CLI volume reset-brick $V0 $H0:$B0/${V0}1 start
+
+EXPECT_WITHIN $PROCESS_DOWN_TIMEOUT 3 count_up_bricks
+EXPECT 1 count_brick_processes
+
+# Negative case with brick killed but volume-id xattr present
+TEST ! $CLI volume reset-brick $V0 $H0:$B0/${V0}1 $H0:$B0/${V0}1 commit
+
+# reset-brick commit force should work and should bring up the brick
+TEST $CLI volume reset-brick $V0 $H0:$B0/${V0}1 $H0:$B0/${V0}1 commit force
+
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 4 count_up_bricks
+EXPECT 1 count_brick_processes
+
+TEST glusterfs --volfile-id=$V1 --volfile-server=$H0 $M1;
+# Create files
+for i in {1..5}
+do
+        echo $i > $M1/file$i.txt
+done
+
+TEST $CLI volume reset-brick $V1 $H0:$B0/${V1}1 start
+
+EXPECT_WITHIN $PROCESS_DOWN_TIMEOUT 3 count_up_bricks
+EXPECT 1 count_brick_processes
+
+# Simulate reset disk
+for i in {1..5}
+do
+        rm -rf $B0/${V1}1/file$i.txt
+done
+
+setfattr -x trusted.glusterfs.volume-id $B0/${V1}1
+setfattr -x trusted.gfid $B0/${V1}1
+
+# Test reset-brick commit. Using CLI_IGNORE_PARTITION since normal CLI  uses
+# the --wignore flag that essentially makes the command act like "commit force"
+TEST $CLI_IGNORE_PARTITION volume reset-brick $V1 $H0:$B0/${V1}1 $H0:$B0/${V1}1 commit
+
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT 4 count_up_bricks
+EXPECT 1 count_brick_processes
diff --git a/tests/include.rc b/tests/include.rc
index 9e6cea4b487..d92361006eb 100644
--- a/tests/include.rc
+++ b/tests/include.rc
@@ -90,7 +90,12 @@ LOGDIR=$(gluster --print-logdir)
 statedumpdir=`gluster --print-statedumpdir`; # Default directory for statedump
 
 CLI="gluster --mode=script --wignore";
-CLI_NO_FORCE="gluster --mode-script";
+CLI_NO_FORCE="gluster --mode=script";
+
+# CLI_IGNORE_PARTITION makes sure that the warning related to bricks being on
+# root partition is ignored while running the command in a "no force" mode
+CLI_IGNORE_PARTITION="gluster --mode=script --wignore-partition"
+
 _GFS () {
 	glusterfs "$@"
 	local mount_ret=$?