gluster rebalance: give option to split the command

the 'gluster volume rebalance <VOLNAME> start' is enhanced with two more options:

* 'gluster volume rebalance <VOLNAME> fix-layout start' (for fixing layout only)
* 'gluster volume rebalance <VOLNAME> migrate-data start' (for migrating data only)

Also the old way of running rebalance in one shot will still work fine

* 'gluster volume rebalance <VOLNAME> start'

Signed-off-by: Amar Tumballi <amar@gluster.com>
Signed-off-by: Anand V. Avati <avati@dev.gluster.com>

BUG: 2258 (enhance gluster volume rebalance)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=2258

6 files changed, 123 insertions, 81 deletions

diff --git a/cli/src/cli-cmd-volume.c b/cli/src/cli-cmd-volume.c
index 9e93d4e973c..a790b326e83 100644
--- a/cli/src/cli-cmd-volume.c
+++ b/cli/src/cli-cmd-volume.c
@@ -476,7 +476,7 @@ cli_cmd_volume_defrag_cbk (struct cli_state *state, struct cli_cmd_word *word,
         if (!dict)
                 goto out;
 
-        if (wordcount != 4) {
+        if (!((wordcount == 4) || (wordcount == 5))) {
                 cli_usage_out (word->pattern);
                 parse_error = 1;
                 goto out;
@@ -486,9 +486,19 @@ cli_cmd_volume_defrag_cbk (struct cli_state *state, struct cli_cmd_word *word,
         if (ret)
                 goto out;
 
-        ret = dict_set_str (dict, "command", (char *)words[3]);
-        if (ret)
-                goto out;
+        if (wordcount == 4) {
+                ret = dict_set_str (dict, "command", (char *)words[3]);
+                if (ret)
+                        goto out;
+        }
+        if (wordcount == 5) {
+                ret = dict_set_str (dict, "start-type", (char *)words[3]);
+                if (ret)
+                        goto out;
+                ret = dict_set_str (dict, "command", (char *)words[4]);
+                if (ret)
+                        goto out;
+        }
 
         proc = &cli_rpc_prog->proctable[GLUSTER_CLI_DEFRAG_VOLUME];
 
@@ -962,7 +972,7 @@ struct cli_cmd volume_cmds[] = {
           cli_cmd_volume_remove_brick_cbk,
           "remove brick from volume <VOLNAME>"},
 
-        { "volume rebalance <VOLNAME> {start|stop|status}",
+        { "volume rebalance <VOLNAME> [fix-layout|migrate-data] {start|stop|status}",
           cli_cmd_volume_defrag_cbk,
           "rebalance operations"},
 
diff --git a/cli/src/cli-rpc-ops.c b/cli/src/cli-rpc-ops.c
index 4e713eef52a..d95c91bbeba 100644
--- a/cli/src/cli-rpc-ops.c
+++ b/cli/src/cli-rpc-ops.c
@@ -803,7 +803,9 @@ gf_cli3_1_defrag_volume_cbk (struct rpc_req *req, struct iovec *iov,
                 volname = local->u.defrag_vol.volname;
                 cmd = local->u.defrag_vol.cmd;
         }
-        if (cmd == GF_DEFRAG_CMD_START) {
+        if ((cmd == GF_DEFRAG_CMD_START) ||
+            (cmd == GF_DEFRAG_CMD_START_LAYOUT_FIX) ||
+            (cmd == GF_DEFRAG_CMD_START_MIGRATE_DATA)) {
                 if (rsp.op_ret && strcmp (rsp.op_errstr, ""))
                         cli_out (rsp.op_errstr);
                 else
@@ -845,21 +847,34 @@ gf_cli3_1_defrag_volume_cbk (struct rpc_req *req, struct iovec *iov,
                                 status = "completed";
                         if (rsp.op_errno == 5)
                                 status = "failed";
+                        if (rsp.op_errno == 6)
+                                status = "step 1: layout fix complete";
+                        if (rsp.op_errno == 7)
+                                status = "step 2: data migration complete";
 
                         if (rsp.files && (rsp.op_errno == 1)) {
                                 cli_out ("rebalance %s: fixed layout %"PRId64,
                                          status, rsp.files);
-                        } else if (rsp.files) {
+                                goto done;
+                        }
+                        if (rsp.files && (rsp.op_errno == 6)) {
+                                cli_out ("rebalance %s: fixed layout %"PRId64,
+                                         status, rsp.files);
+                                goto done;
+                        }
+                        if (rsp.files) {
                                 cli_out ("rebalance %s: rebalanced %"PRId64
                                          " files of size %"PRId64" (total files"
                                          " scanned %"PRId64")", status,
                                          rsp.files, rsp.size, rsp.lookedup_files);
-                        } else {
-                                cli_out ("rebalance %s", status);
+                                goto done;
                         }
+
+                        cli_out ("rebalance %s", status);
                 }
         }
 
+done:
         if (volname)
                 GF_FREE (volname);
 
@@ -1723,15 +1738,28 @@ gf_cli3_1_defrag_volume (call_frame_t *frame, xlator_t *this,
                 goto out;
         }
 
-        if (strncasecmp (cmd_str, "start", 6) == 0) {
+        if (strcmp (cmd_str, "start") == 0) {
                 req.cmd = GF_DEFRAG_CMD_START;
-        } else if (strncasecmp (cmd_str, "stop", 5) == 0) {
+                ret = dict_get_str (dict, "start-type", &cmd_str);
+                if (!ret) {
+                        if (strcmp (cmd_str, "fix-layout") == 0) {
+                                req.cmd = GF_DEFRAG_CMD_START_LAYOUT_FIX;
+                        }
+                        if (strcmp (cmd_str, "migrate-data") == 0) {
+                                req.cmd = GF_DEFRAG_CMD_START_MIGRATE_DATA;
+                        }
+                }
+                goto done;
+        }
+        if (strcmp (cmd_str, "stop") == 0) {
                 req.cmd = GF_DEFRAG_CMD_STOP;
-        } else if (strncasecmp (cmd_str, "status", 7) == 0) {
+                goto done;
+        }
+        if (strcmp (cmd_str, "status") == 0) {
                 req.cmd = GF_DEFRAG_CMD_STATUS;
         }
 
-
+done:
         local = cli_local_get ();
 
         if (local) {
diff --git a/rpc/xdr/src/cli1.h b/rpc/xdr/src/cli1.h
index 2b7c2c65c0f..96b511bfeed 100644
--- a/rpc/xdr/src/cli1.h
+++ b/rpc/xdr/src/cli1.h
@@ -25,9 +25,13 @@
 
 #include "cli1-xdr.h"
 
-#define GF_DEFRAG_CMD_START   1
-#define GF_DEFRAG_CMD_STOP    2
-#define GF_DEFRAG_CMD_STATUS  3
+enum gf_cli_defrag_type {
+        GF_DEFRAG_CMD_START = 1,
+        GF_DEFRAG_CMD_STOP,
+        GF_DEFRAG_CMD_STATUS,
+        GF_DEFRAG_CMD_START_LAYOUT_FIX,
+        GF_DEFRAG_CMD_START_MIGRATE_DATA,
+};
 
 ssize_t
 gf_xdr_serialize_cli_probe_rsp (struct iovec outmsg, void *rsp);
diff --git a/xlators/mgmt/glusterd/src/glusterd-rebalance.c b/xlators/mgmt/glusterd/src/glusterd-rebalance.c
index c1dbead2eed..58c12c19385 100644
--- a/xlators/mgmt/glusterd/src/glusterd-rebalance.c
+++ b/xlators/mgmt/glusterd/src/glusterd-rebalance.c
@@ -184,8 +184,7 @@ gf_glusterd_rebalance_move_data (glusterd_volinfo_t *volinfo, const char *dir)
                 if (!S_ISDIR (stbuf.st_mode))
                         continue;
 
-                ret = gf_glusterd_rebalance_move_data (volinfo,
-                                                       full_path);
+                ret = gf_glusterd_rebalance_move_data (volinfo, full_path);
                 if (ret)
                         break;
         }
@@ -289,32 +288,44 @@ glusterd_defrag_start (void *data)
         sys_lgetxattr (defrag->mount, "trusted.distribute.fix.layout",
                        &value, 128);
 
-        /* root's layout got fixed */
-        defrag->total_files = 1;
+        if ((defrag->cmd == GF_DEFRAG_CMD_START) ||
+            (defrag->cmd == GF_DEFRAG_CMD_START_LAYOUT_FIX)) {
+                /* root's layout got fixed */
+                defrag->total_files = 1;
 
-        /* Step 1: Fix layout of all the directories */
-        ret = gf_glusterd_rebalance_fix_layout (volinfo, defrag->mount);
-        if (ret) {
-                volinfo->defrag_status   = GF_DEFRAG_STATUS_FAILED;
-                goto out;
+                /* Step 1: Fix layout of all the directories */
+                ret = gf_glusterd_rebalance_fix_layout (volinfo, defrag->mount);
+                if (ret) {
+                        volinfo->defrag_status   = GF_DEFRAG_STATUS_FAILED;
+                        goto out;
+                }
+
+                /* Completed first step */
+                volinfo->defrag_status = GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE;
         }
 
-        /* Completed first step */
-        volinfo->defrag_status = GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE;
+        if ((defrag->cmd == GF_DEFRAG_CMD_START) ||
+            (defrag->cmd == GF_DEFRAG_CMD_START_MIGRATE_DATA)) {
+                /* It was used by number of layout fixes on directories */
+                defrag->total_files = 0;
 
-        /* It was used by number of layout fixes on directories */
-        defrag->total_files = 0;
+                volinfo->defrag_status = GF_DEFRAG_STATUS_MIGRATE_DATA_STARTED;
 
-        /* Step 2: Iterate over directories to move data */
-        ret = gf_glusterd_rebalance_move_data (volinfo, defrag->mount);
-        if (ret) {
-                volinfo->defrag_status   = GF_DEFRAG_STATUS_FAILED;
+                /* Step 2: Iterate over directories to move data */
+                ret = gf_glusterd_rebalance_move_data (volinfo, defrag->mount);
+                if (ret) {
+                        volinfo->defrag_status   = GF_DEFRAG_STATUS_FAILED;
+                        goto out;
+                }
+
+                /* Completed second step */
+                volinfo->defrag_status = GF_DEFRAG_STATUS_MIGRATE_DATA_COMPLETE;
         }
 
         /* Completed whole process */
-        if (!ret) {
-                volinfo->defrag_status   = GF_DEFRAG_STATUS_COMPLETE;
-        }
+        if (defrag->cmd == GF_DEFRAG_CMD_START)
+                volinfo->defrag_status = GF_DEFRAG_STATUS_COMPLETE;
+
         volinfo->rebalance_files = defrag->total_files;
         volinfo->rebalance_data  = defrag->total_data;
         volinfo->lookedup_files  = defrag->num_files_lookedup;
@@ -441,6 +452,16 @@ void
 glusterd_rebalance_cmd_attempted_log (int cmd, char *volname)
 {
         switch (cmd) {
+                case GF_DEFRAG_CMD_START_LAYOUT_FIX:
+                        gf_cmd_log ("Volume rebalance"," on volname: %s "
+                                    "cmd: start fix layout , attempted",
+                                    volname);
+                        break;
+                case GF_DEFRAG_CMD_START_MIGRATE_DATA:
+                        gf_cmd_log ("Volume rebalance"," on volname: %s "
+                                    "cmd: start data migrate attempted",
+                                    volname);
+                        break;
                 case GF_DEFRAG_CMD_START:
                         gf_cmd_log ("Volume rebalance"," on volname: %s "
                                     "cmd: start, attempted", volname);
@@ -452,9 +473,9 @@ glusterd_rebalance_cmd_attempted_log (int cmd, char *volname)
                 default:
                         break;
         }
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received rebalance volume %s on %s",
-                (cmd == GF_DEFRAG_CMD_START)?"start":(cmd == GF_DEFRAG_CMD_STOP)?"stop":"status"
-                , volname);
+
+        gf_log ("glusterd", GF_LOG_NORMAL, "Received rebalance volume %d on %s",
+                cmd, volname);
 }
 
 void
@@ -498,7 +519,7 @@ out:
 
 int
 glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr,
-                              size_t len)
+                              size_t len, int cmd)
 {
         int                    ret = -1;
         glusterd_defrag_info_t *defrag =  NULL;
@@ -521,6 +542,8 @@ glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr,
 
         defrag = volinfo->defrag;
 
+        defrag->cmd = cmd;
+
         LOCK_INIT (&defrag->lock);
         snprintf (defrag->mount, 1024, "%s/mount/%s",
                   priv->workdir, volinfo->volname);
@@ -546,7 +569,7 @@ glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr,
                 goto out;
         }
 
-        volinfo->defrag_status = GF_DEFRAG_STATUS_STARTED;
+        volinfo->defrag_status = GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED;
 
         ret = pthread_create (&defrag->th, NULL, glusterd_defrag_start,
                               volinfo);
@@ -608,24 +631,6 @@ glusterd_handle_defrag_volume_v2 (rpcsvc_request_t *req)
                 goto out;
         }
 
-        switch (cli_req.cmd) {
-                case GF_DEFRAG_CMD_START:
-                        gf_cmd_log ("Volume rebalance"," on volname: %s "
-                                    "cmd: start, attempted", cli_req.volname);
-                        break;
-                case GF_DEFRAG_CMD_STOP:
-                        gf_cmd_log ("Volume rebalance"," on volname: %s "
-                                    "cmd: stop, attempted", cli_req.volname);
-                        break;
-                default:
-                        break;
-        }
-
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received rebalance volume %s on %s",
-                ((cli_req.cmd == GF_DEFRAG_CMD_START) ? "start" :
-                 (cli_req.cmd == GF_DEFRAG_CMD_STOP) ? "stop" : "status"),
-                cli_req.volname);
-
         glusterd_rebalance_cmd_attempted_log (cli_req.cmd, cli_req.volname);
 
         rsp.volname = cli_req.volname;
@@ -636,9 +641,13 @@ glusterd_handle_defrag_volume_v2 (rpcsvc_request_t *req)
                                                &volinfo, msg, sizeof (msg));
         if (ret)
                 goto out;
+
         switch (cli_req.cmd) {
         case GF_DEFRAG_CMD_START:
-                ret = glusterd_handle_defrag_start (volinfo, msg, sizeof (msg));
+        case GF_DEFRAG_CMD_START_LAYOUT_FIX:
+        case GF_DEFRAG_CMD_START_MIGRATE_DATA:
+                ret = glusterd_handle_defrag_start (volinfo, msg, sizeof (msg),
+                                                    cli_req.cmd);
                 rsp.op_ret = ret;
                 break;
         case GF_DEFRAG_CMD_STOP:
@@ -684,22 +693,7 @@ glusterd_handle_defrag_volume (rpcsvc_request_t *req)
                 goto out;
         }
 
-        switch (cli_req.cmd) {
-                case GF_DEFRAG_CMD_START:
-                        gf_cmd_log ("Volume rebalance"," on volname: %s "
-                                    "cmd: start, attempted", cli_req.volname);
-                        break;
-                case GF_DEFRAG_CMD_STOP:
-                        gf_cmd_log ("Volume rebalance"," on volname: %s "
-                                    "cmd: stop, attempted", cli_req.volname);
-                        break;
-                default:
-                        break;
-        }
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received rebalance volume %s on %s",
-                ((cli_req.cmd == GF_DEFRAG_CMD_START) ? "start" :
-                 (cli_req.cmd == GF_DEFRAG_CMD_STOP) ? "stop" : "status"),
-                cli_req.volname);
+        glusterd_rebalance_cmd_attempted_log (cli_req.cmd, cli_req.volname);
 
         rsp.volname = cli_req.volname;
         rsp.op_ret = -1;
@@ -710,8 +704,11 @@ glusterd_handle_defrag_volume (rpcsvc_request_t *req)
                 goto out;
         switch (cli_req.cmd) {
         case GF_DEFRAG_CMD_START:
+        case GF_DEFRAG_CMD_START_LAYOUT_FIX:
+        case GF_DEFRAG_CMD_START_MIGRATE_DATA:
         {
-                ret = glusterd_handle_defrag_start (volinfo, msg, sizeof (msg));
+                ret = glusterd_handle_defrag_start (volinfo, msg, sizeof (msg),
+                                                    cli_req.cmd);
                 rsp.op_ret = ret;
                 break;
         }
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index 0c910477635..cb38ddf88f6 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -2240,8 +2240,8 @@ out:
 int
 glusterd_is_defrag_on (glusterd_volinfo_t *volinfo)
 {
-        return ((volinfo->defrag_status == GF_DEFRAG_STATUS_STARTED) ||
-                (volinfo->defrag_status == GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE));
+        return ((volinfo->defrag_status == GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED) ||
+                (volinfo->defrag_status == GF_DEFRAG_STATUS_MIGRATE_DATA_STARTED));
 }
 
 int
diff --git a/xlators/mgmt/glusterd/src/glusterd.h b/xlators/mgmt/glusterd/src/glusterd.h
index 06f35479c35..2b02a513981 100644
--- a/xlators/mgmt/glusterd/src/glusterd.h
+++ b/xlators/mgmt/glusterd/src/glusterd.h
@@ -133,11 +133,13 @@ struct gf_defrag_brickinfo_ {
 
 typedef enum gf_defrag_status_ {
         GF_DEFRAG_STATUS_NOT_STARTED,
-        GF_DEFRAG_STATUS_STARTED,
-        GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE,
+        GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED,
+        GF_DEFRAG_STATUS_MIGRATE_DATA_STARTED,
         GF_DEFRAG_STATUS_STOPED,
         GF_DEFRAG_STATUS_COMPLETE,
         GF_DEFRAG_STATUS_FAILED,
+        GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE,
+        GF_DEFRAG_STATUS_MIGRATE_DATA_COMPLETE,
 } gf_defrag_status_t;
 
 struct glusterd_defrag_info_ {
@@ -145,6 +147,7 @@ struct glusterd_defrag_info_ {
         uint64_t                     total_data;
         uint64_t                     num_files_lookedup;
         gf_lock_t                    lock;
+        int                          cmd;
         pthread_t                    th;
         char                         mount[1024];
         char                         databuf[131072];