glusterd/gverify.sh: Stops session being created with invalid slave details

create force will fail with proper message, if the ip is not
reachable, or is unable to fetch slave details.

Change-Id: I44a3ba777b37702ffd0e48e9cb46c51e293327d4
BUG: 988314
Signed-off-by: Avra Sengupta <asengupt@redhat.com>
Signed-off-by: Venky Shankar <vshankar@redhat.com>
Reviewed-on: http://review.gluster.org/5516
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>

3 files changed, 87 insertions, 19 deletions

diff --git a/extras/hook-scripts/S56glusterd-geo-rep-create-post.sh b/extras/hook-scripts/S56glusterd-geo-rep-create-post.sh
index 71e44b6ed13..1369c22fc1a 100755
--- a/extras/hook-scripts/S56glusterd-geo-rep-create-post.sh
+++ b/extras/hook-scripts/S56glusterd-geo-rep-create-post.sh
@@ -22,6 +22,7 @@ if [ "$val" == "" ]; then
     exit;
 fi
 pub_file=`echo $val`
+pub_file_tmp=`echo $val`_tmp
 
 key=`echo $key_val_pair3 | cut -d '=' -f 1`
 val=`echo $key_val_pair3 | cut -d '=' -f 2`
@@ -34,8 +35,8 @@ fi
 slave_ip=`echo $val`
 
 if [ -f $pub_file ]; then
-    ssh $slave_ip "\rm -rf $pub_file"
-    scp $pub_file $slave_ip:$pub_file &> /dev/null
+    scp $pub_file $slave_ip:$pub_file_tmp
+    ssh $slave_ip "mv $pub_file_tmp $pub_file"
     ssh $slave_ip "gluster system:: copy file /geo-replication/common_secret.pem.pub > /dev/null"
     ssh $slave_ip "gluster system:: execute add_secret_pub > /dev/null"
 fi
diff --git a/geo-replication/src/gverify.sh b/geo-replication/src/gverify.sh
index f10e21ed6e0..49ed2e925ae 100755
--- a/geo-replication/src/gverify.sh
+++ b/geo-replication/src/gverify.sh
@@ -96,6 +96,15 @@ function slave_stats()
 
 function main()
 {
+    log_file=$4
+    > $log_file
+
+    ping -w 5 $2;
+    if [ $? -ne 0 ]; then
+        echo "$2 not reachable." > $log_file
+        exit 1;
+    fi;
+
     ERRORS=0;
     master_data=$(master_stats $1);
     slave_data=$(slave_stats $2 $3);
@@ -104,24 +113,22 @@ function main()
     master_version=$(echo $master_data | cut -f2 -d':');
     slave_version=$(echo $slave_data | cut -f2 -d':');
     slave_no_of_files=$(echo $slave_data | cut -f3 -d':');
-    log_file=$4
-    > $log_file
+    slave_vol_test=$5
 
     if [[ "x$master_size" = "x" || "x$master_version" = "x" || "$master_size" -eq "0" ]]; then
-	echo "Unable to fetch master volume details." > $log_file;
+	echo "Unable to fetch master volume details. Please check the master cluster and master volume." > $log_file;
 	exit 1;
     fi;
 
     if [[ "x$slave_size" = "x" || "x$slave_version" = "x" || "$slave_size" -eq "0" ]]; then
-        ping -w 5 $2;
-        if [ $? -ne 0 ]; then
-            echo "$2 not reachable." > $log_file
-            exit 1;
-        fi;
-	echo "Unable to fetch slave volume details." > $log_file;
+	echo "Unable to fetch slave volume details. Please check the slave cluster and slave volume." > $log_file;
 	exit 1;
     fi;
 
+    if [ ! -z $slave_vol_test ]; then
+        exit $ERRORS;
+    fi
+
     if [ ! $slave_size -ge $(($master_size - $BUFFER_SIZE )) ]; then
 	echo "Total size of master is greater than available size of slave." >> $log_file;
 	ERRORS=$(($ERRORS + 1));
diff --git a/xlators/mgmt/glusterd/src/glusterd-geo-rep.c b/xlators/mgmt/glusterd/src/glusterd-geo-rep.c
index 7ccf641c8b2..46d4d28e601 100644
--- a/xlators/mgmt/glusterd/src/glusterd-geo-rep.c
+++ b/xlators/mgmt/glusterd/src/glusterd-geo-rep.c
@@ -1608,7 +1608,7 @@ out:
 
 static int
 glusterd_verify_slave (char *volname, char *slave_ip, char *slave,
-                       char **op_errstr)
+                       char **op_errstr, int slave_vol_test)
 {
         int32_t         ret     = -1;
         runner_t        runner    = {0,};
@@ -1636,6 +1636,8 @@ glusterd_verify_slave (char *volname, char *slave_ip, char *slave,
         runner_argprintf (&runner, "%s", slave_ip);
         runner_argprintf (&runner, "%s", slave);
         runner_argprintf (&runner, "%s", log_file_path);
+        if (slave_vol_test)
+                runner_add_args  (&runner, "slave_vol_test", NULL);
         runner_redir (&runner, STDOUT_FILENO, RUN_PIPE);
         synclock_unlock (&priv->big_lock);
         ret = runner_run (&runner);
@@ -1661,6 +1663,52 @@ out:
 }
 
 int
+glusterd_validate_slave_ip_vol (dict_t *dict, char *volname,
+                                char *host_uuid, char **op_errstr)
+{
+        int                 ret                       = -1;
+        char                uuid_str [64]             = "";
+        char               *slave                     = NULL;
+        char               *slave_ip                  = NULL;
+        char               *slave_vol                 = NULL;
+
+        uuid_utoa_r (MY_UUID, uuid_str);
+        if (strcmp (uuid_str, host_uuid)) {
+                ret = 0;
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "slave", &slave);
+        if (ret || !slave) {
+                gf_log ("", GF_LOG_ERROR, "Unable to fetch slave from dict");
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_get_slave_info (slave, &slave_ip, &slave_vol);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to fetch slave details.");
+                ret = -1;
+                goto out;
+        }
+
+        /* Checking if slave ip is pingable and slave volume is
+         * valid. Fail even with failure in case of force */
+        ret = glusterd_verify_slave (volname, slave_ip, slave_vol,
+                                     op_errstr, 1);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "%s", *op_errstr);
+                ret = -1;
+                goto out;
+        }
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
 glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr)
 {
         char               *down_peerstr              = NULL;
@@ -1748,8 +1796,10 @@ glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr)
                                 volinfo->volname, slave);
                 }
 
+                /* Checking if slave vol is empty, and if it has enough memory
+                 * available, and bypass in case of force */
                 ret = glusterd_verify_slave (volname, slave_ip, slave_vol,
-                                             op_errstr);
+                                             op_errstr, 0);
                 if ((ret) && !is_force) {
                         gf_log ("", GF_LOG_ERROR,
                                 "%s is not a valid slave volume. Error: %s",
@@ -1798,6 +1848,13 @@ glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr)
 
         ret = glusterd_get_statefile_name (volinfo, slave, conf_path, &statefile);
         if (ret) {
+                ret = glusterd_validate_slave_ip_vol (dict, volname,
+                                                      host_uuid, op_errstr);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Slave validation failed.");
+                        goto out;
+                }
+
                 if (!strstr(slave, "::"))
                         snprintf (errmsg, sizeof (errmsg),
                                   "%s is not a valid slave url.", slave);
@@ -1805,6 +1862,7 @@ glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr)
                         snprintf (errmsg, sizeof (errmsg), "Please check gsync "
                                   "config file. Unable to get statefile's name");
                 gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                ret = -1;
                 goto out;
         }
 
@@ -1889,7 +1947,7 @@ glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr)
         }
 
         ret = glusterd_op_gsync_args_get (dict, op_errstr,
-                                          &volname, &slave, NULL);
+                                          &volname, &slave, &host_uuid);
         if (ret)
                 goto out;
 
@@ -1916,6 +1974,13 @@ glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr)
 
         ret = glusterd_get_statefile_name (volinfo, slave, conf_path, &statefile);
         if (ret) {
+                ret = glusterd_validate_slave_ip_vol (dict, volname,
+                                                      host_uuid, op_errstr);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Slave validation failed.");
+                        goto out;
+                }
+
                 if (!strstr(slave, "::"))
                         snprintf (errmsg, sizeof (errmsg),
                                   "%s is not a valid slave url.", slave);
@@ -1958,11 +2023,6 @@ glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr)
         /* Check if all peers that are a part of the volume are up or not */
         if ((type == GF_GSYNC_OPTION_TYPE_DELETE) ||
             ((type == GF_GSYNC_OPTION_TYPE_STOP) && !is_force)) {
-                ret = dict_get_str (dict, "host-uuid", &host_uuid);
-                if (ret < 0)
-                        goto out;
-
-                uuid_utoa_r (MY_UUID, uuid_str);
                 if (!strcmp (uuid_str, host_uuid)) {
                         ret = glusterd_are_vol_all_peers_up (volinfo,
                                                              &conf->peers,