summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorDan Lambright <dlambrig@redhat.com>2015-06-23 16:35:03 -0400
committerDan Lambright <dlambrig@redhat.com>2015-06-26 17:10:56 -0700
commit875aa01ec80e56d85d0bc6028c6f1417f6ab140f (patch)
treee5d57c306ba51ec0596f83fbab82364274ecdd1d
parent26ef697318a7fec5ed82b000e3be4e30cfb16b50 (diff)
cluster/tier: stop tier migration after graph switch
On a graph switch, a new xlator and private structures are created. The tier migration daemon must stop using the old xlator and private structures and begin using the new ones. Otherwise, when RPCs arrive (such as counter queries from glusterd), the new xlator will be consulted but it will not have up to date information. The fix detects a graph switch and exits the daemon in this case. Typical graph switches for the tier case would be turning off performance translators. Change-Id: Ibfbd4720dc82ea179b77c81b8f534abced21e3c8 BUG: 1226005 Signed-off-by: Dan Lambright <dlambrig@redhat.com> Reviewed-on: http://review.gluster.org/11372
-rwxr-xr-xtests/basic/tier/tier.t21
-rw-r--r--xlators/cluster/dht/src/tier.c16
2 files changed, 34 insertions, 3 deletions
diff --git a/tests/basic/tier/tier.t b/tests/basic/tier/tier.t
index b3daae35843..ee730e727bc 100755
--- a/tests/basic/tier/tier.t
+++ b/tests/basic/tier/tier.t
@@ -52,6 +52,15 @@ function confirm_vol_stopped {
fi
}
+function check_counters_nonzero {
+ $CLI volume rebalance $V0 tier status | grep ' 0 '
+ if [ $? == 0 ]; then
+ echo "1"
+ else
+ echo "0"
+ fi
+}
+
DEMOTE_TIMEOUT=12
PROMOTE_TIMEOUT=5
MIGRATION_TIMEOUT=10
@@ -62,12 +71,17 @@ TEST pidof glusterd
TEST $CLI volume create $V0 replica 2 $H0:$B0/${V0}{0..$LAST_BRICK}
# testing bug 1215122, ie should fail if replica count and bricks are not compatible.
+
TEST ! $CLI volume attach-tier $V0 replica 5 $H0:$B0/${V0}$CACHE_BRICK_FIRST $H0:$B0/${V0}$CACHE_BRICK_LAST
TEST $CLI volume start $V0
+# The following two commands instigate a graph switch. Do them
+# before attaching the tier. If done on a tiered volume the rebalance
+# daemon will terminate and must be restarted manually.
TEST $CLI volume set $V0 performance.quick-read off
TEST $CLI volume set $V0 performance.io-cache off
+
TEST $CLI volume set $V0 features.ctr-enabled on
#Not a tier volume
@@ -78,6 +92,8 @@ TEST ! $CLI volume detach-tier $V0 commit force
TEST $CLI volume attach-tier $V0 replica 2 $H0:$B0/${V0}$CACHE_BRICK_FIRST $H0:$B0/${V0}$CACHE_BRICK_LAST
+$CLI volume rebalance $V0 tier status
+
#Tier options expect non-negative value
TEST ! $CLI volume set $V0 cluster.tier-promote-frequency -1
@@ -128,13 +144,12 @@ sleep 5
EXPECT_WITHIN $PROMOTE_TIMEOUT "0" file_on_fast_tier d1/data2.txt
EXPECT_WITHIN $PROMOTE_TIMEOUT "0" file_on_fast_tier d1/data3.txt
+EXPECT "0" check_counters_nonzero
+
# stop gluster, when it comes back info file should have tiered volume
killall glusterd
TEST glusterd
-# Test rebalance commands
-TEST $CLI volume rebalance $V0 tier status
-
TEST $CLI volume detach-tier $V0 start
TEST $CLI volume detach-tier $V0 commit force
diff --git a/xlators/cluster/dht/src/tier.c b/xlators/cluster/dht/src/tier.c
index c8c6ad94c7e..650414c50d9 100644
--- a/xlators/cluster/dht/src/tier.c
+++ b/xlators/cluster/dht/src/tier.c
@@ -776,6 +776,8 @@ tier_start (xlator_t *this, gf_defrag_info_t *defrag)
pthread_t demote_thread;
gf_boolean_t is_promotion_triggered = _gf_false;
gf_boolean_t is_demotion_triggered = _gf_false;
+ xlator_t *any = NULL;
+ xlator_t *xlator = NULL;
conf = this->private;
@@ -798,6 +800,20 @@ tier_start (xlator_t *this, gf_defrag_info_t *defrag)
while (1) {
+ /*
+ * Check if a graph switch occured. If so, stop migration
+ * thread. It will need to be restarted manually.
+ */
+ any = THIS->ctx->active->first;
+ xlator = xlator_search_by_name(any, this->name);
+
+ if (xlator != this) {
+ gf_msg (this->name, GF_LOG_INFO, 0,
+ DHT_MSG_LOG_TIER_STATUS,
+ "Detected graph switch. Exiting migration daemon.");
+ goto out;
+ }
+
sleep(1);
if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {