cluster/dht : Do not migrate files with POSIX locks held

dht_migrate_file does not migrate file locks to the dst file.
Any locks held on the source file are lost once the migration
is complete. This issue is magnified in the case of a tier volume
as file migrations occur more frequently and repeatedly as compared
to a DHT rebalance.

The fix makes 2 changes:
1. Before starting the actual migration process, check if there are
 any locks held on the file. If yes, do not migrate the file.
2. The rebalance process tries to lock on the entire file just before
 moving into the Phase 2 of the file migration. If the lock acquisition
fails, the file migration does not proceed.
If the lock is granted, the file migration proceeds.

This still leaves a small window where conflicting locks can be granted to
different clients. If client1 requests a lock on the src file just after
it is converted to a linkto file and client2 requests a lock on the dst
data file, they will both be granted, but all FOPs will be redirected
to the dst data file. This issue will be taken up in a subsequent patch.

Change-Id: I8c895fc3cced50dd2894259d40a827c7b43d58ac
BUG: 1271148
Signed-off-by: N Balachandran <nbalacha@redhat.com>
Reviewed-on: http://review.gluster.org/12347
Tested-by: NetBSD Build System <jenkins@build.gluster.org>
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Dan Lambright <dlambrig@redhat.com>
Tested-by: Dan Lambright <dlambrig@redhat.com>

2 files changed, 187 insertions, 0 deletions

diff --git a/tests/basic/tier/file_lock.c b/tests/basic/tier/file_lock.c
new file mode 100644
index 00000000000..730cca92e42
--- /dev/null
+++ b/tests/basic/tier/file_lock.c
@@ -0,0 +1,75 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <fcntl.h>
+
+
+void usage (void)
+{
+
+        printf ("Usage: testlock <filepath> [R|W]\n");
+        return;
+}
+
+
+int main (int argc, char *argv[])
+{
+        char *file_path = NULL;
+        int fd = -1;
+        struct flock lock = {0};
+        int ret = -1;
+        int c = 0;
+
+        if (argc != 3) {
+                usage ();
+                exit (1);
+        }
+
+        file_path = argv[1];
+        fd = open (file_path, O_RDWR);
+
+        if (-1 == fd) {
+                printf ("Failed to open file %s. %m\n", file_path);
+                exit (1);
+        }
+
+        /* TODO: Check for invalid input*/
+
+        if (!strcmp (argv[2], "W")) {
+                lock.l_type = F_WRLCK;
+                printf("Taking write lock\n");
+
+        } else {
+                lock.l_type = F_RDLCK;
+                printf("Taking read lock\n");
+        }
+
+        lock.l_whence = SEEK_SET;
+        lock.l_start = 0;
+        lock.l_len = 0;
+        lock.l_pid = getpid ();
+
+
+        printf ("Acquiring lock on %s\n", file_path);
+        ret = fcntl (fd, F_SETLK, &lock);
+        if (ret) {
+                printf ("Failed to acquire lock on %s (%m)\n", file_path);
+                close (fd);
+                exit (1);
+        }
+
+        sleep(10);
+
+        /*Unlock*/
+
+        printf ("Releasing lock on %s\n", file_path);
+        lock.l_type = F_UNLCK;
+        ret = fcntl (fd, F_SETLK, &lock);
+        if (ret) {
+                printf ("Failed to release lock on %s (%m)\n", file_path);
+        }
+
+        close (fd);
+        return ret;
+
+}
diff --git a/tests/basic/tier/locked_file_migration.t b/tests/basic/tier/locked_file_migration.t
new file mode 100755
index 00000000000..c3ba1b27749
--- /dev/null
+++ b/tests/basic/tier/locked_file_migration.t
@@ -0,0 +1,112 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+
+
+NUM_BRICKS=3
+DEMOTE_FREQ=7
+PROMOTE_FREQ=30
+DEMOTE_TIMEOUT=15
+
+TEST_STR="Testing write and truncate fops on tier migration"
+
+function is_sticky_set () {
+        echo $1
+        if [ -k $1 ];
+        then
+                echo "yes"
+        else
+                echo "no"
+        fi
+}
+
+function check_counters {
+    index=0
+    ret=0
+    rm -f /tmp/tc*.txt
+    echo "0" > /tmp/tc2.txt
+    $CLI volume rebalance $V0 tier status | grep localhost > /tmp/tc.txt
+
+    promote=`cat /tmp/tc.txt |awk '{print $2}'`
+    demote=`cat /tmp/tc.txt |awk '{print $3}'`
+    if [ "${promote}" != "${1}" ]; then
+         echo "1" > /tmp/tc2.txt
+
+    elif [ "${demote}" != "${2}" ]; then
+         echo "2" > /tmp/tc2.txt
+    fi
+
+    # temporarily disable non-Linux tests.
+    case $OSTYPE in
+        NetBSD | FreeBSD | Darwin)
+            echo "0" > /tmp/tc2.txt
+            ;;
+    esac
+    cat /tmp/tc2.txt
+}
+
+
+# Creates a tiered volume with pure distribute hot and cold tiers
+# Both hot and cold tiers will have an equal number of bricks.
+
+function create_dist_tier_vol () {
+        mkdir $B0/cold
+        mkdir $B0/hot
+        TEST $CLI volume create $V0 $H0:$B0/cold/${V0}{0..$1}
+        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 start $V0
+        TEST $CLI volume attach-tier $V0 $H0:$B0/hot/${V0}{0..$1}
+        TEST $CLI volume set $V0 cluster.tier-demote-frequency $DEMOTE_FREQ
+        TEST $CLI volume set $V0 cluster.tier-promote-frequency $PROMOTE_FREQ
+
+#We don't want promotes to happen in this test
+        TEST $CLI volume set $V0 cluster.read-freq-threshold 10
+        TEST $CLI volume set $V0 cluster.write-freq-threshold 10
+}
+
+
+cleanup;
+
+#Basic checks
+TEST glusterd
+TEST pidof glusterd
+TEST $CLI volume info
+
+
+# Create and start a tiered volume
+create_dist_tier_vol $NUM_BRICKS
+
+# Mount FUSE
+TEST glusterfs -s $H0 --volfile-id $V0 $M0
+
+TEST mkdir $M0/dir1
+build_tester $(dirname $0)/file_lock.c -o file_lock
+cp $(dirname $0)/file_lock $M0/file_lock
+
+# The files will be created on the hot tier
+touch $M0/dir1/FILE1
+touch $M0/dir1/FILE2
+
+# For FILE1, take a POSIX write lock on the entire file.
+# Don't take a lock on FILE2
+
+./file_lock $M0/dir1/FILE1 W &
+
+sleep $DEMOTE_FREQ
+
+# Wait for the tier process to demote the file
+# Only FILE2 and file_lock should be demoted
+# FILE1 should be skipped because of the lock held
+# on it
+
+EXPECT_WITHIN $DEMOTE_TIMEOUT "0" check_counters 0 2
+
+sleep 10
+
+rm $(dirname $0)/file_lock
+
+cleanup;
+