cluster/dht: avoid overwriting client writes during migration

For more details on this issue see
https://github.com/gluster/glusterfs/issues/308

Solution:
This is a restrictive solution where a file will not be migrated
if a client writes to it during the migration. This does not
check if the writes from the rebalance and the client actually
do overlap.

If dht_writev_cbk finds that the file is being migrated (PHASE1)
it will set an xattr on the destination file indicating the file
was updated by a non-rebalance client.
Rebalance checks if any other client has written to the dst file
and aborts the file migration if it finds the xattr.

updates gluster/glusterfs#308

Change-Id: I73aec28bc9dbb8da57c7425ec88c6b6af0fbc9dd
Signed-off-by: Susant Palai <spalai@redhat.com>
Signed-off-by: Raghavendra G <rgowdapp@redhat.com>
Signed-off-by: N Balachandran <nbalacha@redhat.com>

2 files changed, 51 insertions, 0 deletions

diff --git a/tests/basic/distribute/force-migration.t b/tests/basic/distribute/force-migration.t
new file mode 100644
index 00000000000..f6c4997a505
--- /dev/null
+++ b/tests/basic/distribute/force-migration.t
@@ -0,0 +1,50 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+
+#This tests checks if the file migration fails with force-migration
+#option set to off.
+
+cleanup;
+
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume create $V0 $H0:$B0/${V0}0 $H0:$B0/${V0}1
+TEST $CLI volume set $V0 performance.quick-read off
+TEST $CLI volume set $V0 performance.io-cache off
+TEST $CLI volume set $V0 performance.write-behind off
+TEST $CLI volume set $V0 performance.stat-prefetch off
+TEST $CLI volume set $V0 performance.read-ahead off
+TEST $CLI volume start $V0
+TEST glusterfs --volfile-id=/$V0 --volfile-server=$H0 $M0 --attribute-timeout=0 --entry-timeout=0
+TEST touch $M0/file
+#This rename creates a link file for tile in the other brick.
+TEST mv $M0/file $M0/tile
+#Lets keep writing to the file which will have a open fd
+dd if=/dev/zero of=$M0/tile bs=1b &
+bg_pid=$!
+#Now rebalance will try to skip the file
+TEST $CLI volume set $V0 force-migration off
+TEST $CLI volume rebalance $V0 start force
+EXPECT_WITHIN $REBALANCE_TIMEOUT "completed" rebalance_status_field $V0
+skippedcount=`gluster v rebalance $V0 status | awk 'NR==3{print $6}'`
+TEST [[ $skippedcount -eq 1 ]]
+#file should be migrated now
+TEST $CLI volume set $V0 force-migration on
+TEST $CLI volume rebalance $V0 start force
+EXPECT_WITHIN $REBALANCE_TIMEOUT "completed" rebalance_status_field $V0
+skippedcount=`gluster v rebalance $V0 status | awk 'NR==3{print $6}'`
+rebalancedcount=`gluster v rebalance $V0 status | awk 'NR==3{print $2}'`
+TEST [[ $skippedcount -eq 0 ]]
+TEST [[ $rebalancedcount -eq 1 ]]
+kill -9 $bg_pid > /dev/null 2>&1
+wait > /dev/null 2>&1
+cleanup
+#Bad test because we are not sure writes are happening at the time of
+#rebalance. We need to write a test case which makes sure client
+#writes happen during rebalance. One way would be to set S+T bits on
+#src and write to file from client and then start rebalance. Currently
+#marking this as bad test.
+#G_TESTDEF_TEST_STATUS_CENTOS6=BAD_TEST,BUG=000000
+
diff --git a/tests/basic/tier/fops-during-migration.t b/tests/basic/tier/fops-during-migration.t
index c85b4e337cc..458c01e93c5 100755
--- a/tests/basic/tier/fops-during-migration.t
+++ b/tests/basic/tier/fops-during-migration.t
@@ -22,6 +22,7 @@ function create_dist_tier_vol () {
         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
+        TEST $CLI volume set $V0 cluster.force-migration on
         TEST $CLI volume start $V0
         TEST $CLI volume tier $V0 attach $H0:$B0/hot/${V0}{0..$1}
         TEST $CLI volume set $V0 cluster.tier-demote-frequency $DEMOTE_FREQ