dht: make lookup-unhashed=auto do something actually useful

The key concept here is to determine whether a directory is "clean" by
comparing its last-known-good topology to the current one for the
volume.  These are stored as "commit hashes" on the directory and the
volume root respectively.  The volume's commit hash changes whenever a
brick is added or removed, and a fix-layout is done.  A directory's
commit hash changes only when a full rebalance (not just fix-layout)
is done on it.  If all bricks are present and have a directory
commit hash that matches the volume commit hash, then we can assume
that every file is in its "proper" place. Therefore, if we look for
a file in that proper place and don't find it, we can assume it's not
on any other subvolume and *safely* skip the global (broadcast to all)
lookup.

Change-Id: Id6ce4593ba1f7daffa74cfab591cb45960629ae3
BUG: 1219637
Signed-off-by: Jeff Darcy <jdarcy@redhat.com>
Signed-off-by: Shyam <srangana@redhat.com>
Reviewed-on: http://review.gluster.org/7702
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Tested-by: NetBSD Build System
Reviewed-by: Vijay Bellur <vbellur@redhat.com>

1 files changed, 99 insertions, 0 deletions

diff --git a/tests/features/unhashed-auto.t b/tests/features/unhashed-auto.t
new file mode 100755
index 00000000000..97663c20e10
--- /dev/null
+++ b/tests/features/unhashed-auto.t
@@ -0,0 +1,99 @@
+#!/bin/bash
+
+. $(dirname $0)/../include.rc
+. $(dirname $0)/../volume.rc
+. $(dirname $0)/../dht.rc
+
+NFILES=100
+
+touch_files () {
+	for i in $(seq 1 $NFILES); do
+		touch $(printf $M0/dir/file%02d $i)
+	done
+}
+
+count_files () {
+	found=0
+	for i in $(seq 1 $NFILES); do
+		if [ -f $(printf $M0/dir/file%02d $i) ]; then
+			found=$((found+1))
+		fi
+	done
+	echo "found $found files" > /dev/tty
+	echo $found
+}
+
+wait_for_rebalance () {
+	while true; do
+		tmp=$(rebalance_completed)
+		if [ $tmp -eq 1 ]; then
+			sleep 1
+		else
+			break
+		fi
+	done
+}
+
+get_xattr () {
+	cmd="getfattr --absolute-names --only-values -n trusted.glusterfs.dht"
+	$cmd $1 | od -tx1 -An | tr -d ' '
+}
+
+cleanup
+
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume info
+
+TEST $CLI volume create $V0 $H0:$B0/${V0}{1,2}
+EXPECT "$V0" volinfo_field $V0 'Volume Name'
+EXPECT 'Created' volinfo_field $V0 'Status'
+
+TEST $CLI volume set $V0 cluster.lookup-unhashed auto
+
+TEST $CLI volume start $V0
+EXPECT 'Started' volinfo_field $V0 'Status'
+
+# Create some files for later tests.
+TEST $GFS -s $H0 --volfile-id $V0 $M0
+TEST mkdir $M0/dir
+TEST touch_files
+TEST umount $M0
+
+# Add a brick and do the fix-layout part of rebalance to update directory layouts
+# (including their directory commit hashes).
+TEST $CLI volume add-brick $V0 $H0:$B0/${V0}3
+EXPECT '3' brick_count $V0
+TEST $CLI volume rebalance $V0 fix-layout start
+TEST wait_for_rebalance
+
+# Now for the sneaky part.  *Undo* the part of rebalance that updated the volume
+# commit hash, forcing a false match between that and the directory commit hashes.
+TEST setfattr -x trusted.glusterfs.dht.commithash $B0/${V0}1
+TEST setfattr -x trusted.glusterfs.dht.commithash $B0/${V0}2
+TEST setfattr -x trusted.glusterfs.dht.commithash $B0/${V0}3
+
+# Mount and check that we do *not* see all of the files.  This indicates that we
+# correctly skipped the broadcast lookup that would have found them.
+TEST $GFS -s $H0 --volfile-id $V0 $M0
+TEST [ $(count_files) -ne 100 ]
+TEST umount $M0
+
+# Do the fix-layout again to generate a new volume commit hash.
+TEST $CLI volume rebalance $V0 fix-layout start
+TEST wait_for_rebalance
+
+# Mount and check that we *do* see all of the files.  This indicates that we saw
+# the mismatch and did the broadcast lookup this time.
+TEST $GFS -s $H0 --volfile-id $V0 $M0
+TEST [ $(count_files) -eq 100 ]
+TEST umount $M0
+
+# Do a *full* rebalance and verify that the directory commit hash changed.
+old_val=$(get_xattr $B0/${V0}1/dir)
+TEST $CLI volume rebalance $V0 start
+TEST wait_for_rebalance
+new_val=$(get_xattr $B0/${V0}1/dir)
+TEST [ ! x"$old_val" = x"$new_val" ]
+
+cleanup