cluster/dht: Correct min_free_disk behaviour

Problem:
Files were being created in subvol which had less than
min_free_disk available even in the cases where other
subvols with more space were available.

Solution:
Changed the logic to look for subvol which has more
space available.
In cases where all the subvols have lesser than
Min_free_disk available , the one with max space and
atleast one inode is available.

Known Issue: Cannot ensure that first file that is
created right after min-free-value is crossed on a
brick will get created in other brick because disk
usage stat takes some time to update in glusterprocess.
Will fix that as part of another bug.

Change-Id: If3ae0bf5a44f8739ce35b3ee3f191009ddd44455
BUG: 858488
Signed-off-by: Raghavendra Talur <rtalur@redhat.com>
Reviewed-on: http://review.gluster.org/4420
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>

4 files changed, 218 insertions, 27 deletions

diff --git a/tests/bugs/bug-858488-min-free-disk.t b/tests/bugs/bug-858488-min-free-disk.t
new file mode 100644
index 00000000000..43ef1496ba1
--- /dev/null
+++ b/tests/bugs/bug-858488-min-free-disk.t
@@ -0,0 +1,114 @@
+#!/bin/bash
+
+. $(dirname $0)/../include.rc
+. $(dirname $0)/../volume.rc
+
+cleanup;
+
+function pidgrep()
+{
+    ps ax | grep "$1" | awk '{print $1}' | head -1
+}
+
+## Start glusterd
+TEST glusterd;
+TEST pidof glusterd;
+TEST $CLI volume info;
+
+## Lets create partitions for bricks
+TEST truncate -s 100M $B0/brick1
+TEST truncate -s 200M $B0/brick2
+TEST LO1=`losetup --find --show $B0/brick1`
+TEST mkfs.xfs $LO1
+TEST LO2=`losetup --find --show $B0/brick2`
+TEST mkfs.xfs $LO2
+TEST mkdir -p $B0/${V0}1 $B0/${V0}2
+TEST mount -t xfs $LO1 $B0/${V0}1
+TEST mount -t xfs $LO2 $B0/${V0}2
+
+
+## Lets create volume
+TEST $CLI volume create $V0 $H0:$B0/${V0}{1,2};
+
+## Verify volume is created
+EXPECT "$V0" volinfo_field $V0 'Volume Name';
+EXPECT 'Created' volinfo_field $V0 'Status';
+
+## Start volume and verify
+TEST $CLI volume start $V0;
+EXPECT 'Started' volinfo_field $V0 'Status';
+TEST glusterfs -s $H0 --volfile-id=$V0 --acl $M0
+MOUNT_PID=`ps ax |grep "glusterfs -s $H0 --volfile-id=$V0 --acl $M0" | awk '{print $1}' | head -1`
+## Real test starts here
+## ----------------------------------------------------------------------------
+
+MINFREEDISKVALUE=90
+
+## Set min free disk to MINFREEDISKVALUE percent
+TEST $CLI volume set $V0 cluster.min-free-disk $MINFREEDISKVALUE
+
+## We need to have file name to brick map based on hash.
+## We will use this info in test case 0.
+i=1
+CONTINUE=2
+BRICK1FILE=0
+BRICK2FILE=0
+while [[ $CONTINUE -ne 0 ]]
+do
+        dd if=/dev/zero of=$M0/file$i.data bs=1024 count=1024 1>/dev/null 2>&1
+
+        if  [[ -e  $B0/${V0}1/file$i.data &&  $BRICK1FILE = "0" ]]
+        then
+                BRICK1FILE=file$i.data
+                CONTINUE=$CONTINUE-1
+        fi
+
+        if [[ -e  $B0/${V0}2/file$i.data &&  $BRICK2FILE = "0" ]]
+        then
+                BRICK2FILE=file$i.data
+                CONTINUE=$CONTINUE-1
+        fi
+
+        rm $M0/file$i.data
+        let i++
+done
+
+
+## Bring free space on one of the bricks to less than minfree value by
+## creating one big file.
+dd if=/dev/zero of=$M0/fillonebrick.data bs=1024 count=25600 1>/dev/null 2>&1
+
+#Lets find out where it was created
+if [ -f $B0/${V0}1/fillonebrick.data ]
+then
+        FILETOCREATE=$BRICK1FILE
+        OTHERBRICK=$B0/${V0}2
+else
+        FILETOCREATE=$BRICK2FILE
+        OTHERBRICK=$B0/${V0}1
+fi
+
+##--------------------------------TEST CASE 0-----------------------------------
+## If we try to create a file which should go into full brick as per hash, it
+## should go into the other brick instead.
+
+## Before that let us create files just to make gluster refresh the stat
+## Using touch so it should not change the disk usage stats
+for k in {1..20};
+do
+        touch $M0/dummyfile$k
+done
+
+dd if=/dev/zero of=$M0/$FILETOCREATE bs=1024 count=2048 1>/dev/null 2>&1
+TEST [ -e $OTHERBRICK/$FILETOCREATE ]
+
+## Done testing, lets clean up
+EXPECT "$MOUNT_PID" pidgrep $MOUNT_PID
+TEST rm -rf $M0/*
+
+## Finish up
+TEST $CLI volume stop $V0;
+EXPECT 'Stopped' volinfo_field $V0 'Status';
+$CLI volume delete $V0;
+
+cleanup;
diff --git a/tests/include.rc b/tests/include.rc
index 1d1386f9826..5d9c96cea0e 100644
--- a/tests/include.rc
+++ b/tests/include.rc
@@ -168,6 +168,21 @@ function cleanup()
 {
 	killall -15 glusterfs glusterfsd glusterd 2>/dev/null || true;
 	killall -9 glusterfs glusterfsd glusterd 2>/dev/null || true;
+
+        MOUNTPOINTS=`mount | grep "$B0/" | awk '{print $3}'`
+        for m in $MOUNTPOINTS;
+        do
+                umount $m
+        done
+
+
+        LOOPDEVICES=`losetup -a | grep "$B0/" | awk '{print $1}' | tr -d :`
+        for l in $LOOPDEVICES;
+        do
+                losetup -d $l
+        done
+
+
 	rm -rf /var/lib/glusterd/* $B0/* /etc/glusterd/*;
 
 	umount -l $M0 2>/dev/null || true;
diff --git a/xlators/cluster/dht/src/dht-common.h b/xlators/cluster/dht/src/dht-common.h
index 764b37ac4a0..0dd654650c0 100644
--- a/xlators/cluster/dht/src/dht-common.h
+++ b/xlators/cluster/dht/src/dht-common.h
@@ -724,4 +724,8 @@ int
 dht_dir_has_layout (dict_t *xattr);
 gf_boolean_t
 dht_is_subvol_in_layout (dht_layout_t *layout, xlator_t *xlator);
+xlator_t *
+dht_subvol_with_free_space_inodes (xlator_t *this, xlator_t *subvol);
+xlator_t *
+dht_subvol_maxspace_nonzeroinode (xlator_t *this, xlator_t *subvol);
 #endif/* _DHT_H */
diff --git a/xlators/cluster/dht/src/dht-diskusage.c b/xlators/cluster/dht/src/dht-diskusage.c
index 52ea3a32aca..0c87f4a647c 100644
--- a/xlators/cluster/dht/src/dht-diskusage.c
+++ b/xlators/cluster/dht/src/dht-diskusage.c
@@ -248,12 +248,11 @@ dht_is_subvol_filled (xlator_t *this, xlator_t *subvol)
 	return is_subvol_filled;
 }
 
+
+/*Get the best subvolume to create the file in*/
 xlator_t *
 dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol)
 {
-	int         i = 0;
-	double      max = 0;
-	double      max_inodes = 0;
 	xlator_t   *avail_subvol = NULL;
 	dht_conf_t *conf = NULL;
 
@@ -261,37 +260,96 @@ dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol)
 
 	LOCK (&conf->subvolume_lock);
 	{
-		for (i = 0; i < conf->subvolume_cnt; i++) {
-			if (conf->disk_unit == 'p') {
-				if ((conf->du_stats[i].avail_percent > max)
-				    && (conf->du_stats[i].avail_inodes > max_inodes)) {
-					max = conf->du_stats[i].avail_percent;
-					max_inodes = conf->du_stats[i].avail_inodes;
-					avail_subvol = conf->subvolumes[i];
-				}
-			} else {
-				if ((conf->du_stats[i].avail_space > max)
-				    && (conf->du_stats[i].avail_inodes > max_inodes)) {
-					max = conf->du_stats[i].avail_space;
-					max_inodes = conf->du_stats[i].avail_inodes;
-					avail_subvol = conf->subvolumes[i];
-				}
+                avail_subvol = dht_subvol_with_free_space_inodes(this, subvol);
+                if(!avail_subvol)
+                {
+                        avail_subvol = dht_subvol_maxspace_nonzeroinode(this,
+                                                                        subvol);
+                }
 
-			}
-		}
 	}
 	UNLOCK (&conf->subvolume_lock);
 
 	if (!avail_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume has enough free space and inodes to create");
+		gf_log (this->name,
+                        GF_LOG_DEBUG,
+			"no subvolume has enough free space and/or inodes\
+                         to create");
+                avail_subvol = subvol;
 	}
 
-	if ((max < conf->min_free_disk) && (max_inodes < conf->min_free_inodes))
-		avail_subvol = subvol;
-
-	if (!avail_subvol)
-		avail_subvol = subvol;
 
 	return avail_subvol;
 }
+
+/*Get subvolume which has both space and inodes more than the min criteria*/
+xlator_t *
+dht_subvol_with_free_space_inodes(xlator_t *this, xlator_t *subvol)
+{
+        int i = 0;
+        double max = 0;
+        double max_inodes = 0;
+
+        xlator_t *avail_subvol = NULL;
+        dht_conf_t *conf = NULL;
+
+        conf = this->private;
+
+        for(i=0; i < conf->subvolume_cnt; i++) {
+                if ((conf->disk_unit == 'p') &&
+                    (conf->du_stats[i].avail_percent > conf->min_free_disk) &&
+                    (conf->du_stats[i].avail_inodes  > conf->min_free_inodes)) {
+                        if ((conf->du_stats[i].avail_inodes > max_inodes) ||
+                            (conf->du_stats[i].avail_percent > max)) {
+                                max = conf->du_stats[i].avail_percent;
+                                max_inodes = conf->du_stats[i].avail_inodes;
+                                avail_subvol = conf->subvolumes[i];
+                        }
+                }
+
+                if ((conf->disk_unit != 'p') &&
+                    (conf->du_stats[i].avail_space > conf->min_free_disk) &&
+                    (conf->du_stats[i].avail_inodes  > conf->min_free_inodes)) {
+                        if ((conf->du_stats[i].avail_inodes > max_inodes) ||
+                            (conf->du_stats[i].avail_space > max)) {
+                                max = conf->du_stats[i].avail_space;
+                                max_inodes = conf->du_stats[i].avail_inodes;
+                                avail_subvol = conf->subvolumes[i];
+                        }
+                }
+        }
+
+        return avail_subvol;
+}
+
+
+/* Get subvol which has atleast one inode and maximum space */
+xlator_t *
+dht_subvol_maxspace_nonzeroinode (xlator_t *this, xlator_t *subvol)
+{
+        int         i = 0;
+        double      max = 0;
+
+        xlator_t   *avail_subvol = NULL;
+        dht_conf_t *conf = NULL;
+
+        conf = this->private;
+
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->disk_unit == 'p') {
+                        if ((conf->du_stats[i].avail_percent > max)
+                            && (conf->du_stats[i].avail_inodes > 0 )) {
+                                max = conf->du_stats[i].avail_percent;
+                                avail_subvol = conf->subvolumes[i];
+                        }
+               } else {
+                         if ((conf->du_stats[i].avail_space > max)
+                            && (conf->du_stats[i].avail_inodes > 0)) {
+                                 max = conf->du_stats[i].avail_space;
+                                 avail_subvol = conf->subvolumes[i];
+                         }
+               }
+        }
+
+        return avail_subvol;
+}