cli : new volume statedump command

Changes:
        1. Add a new 'volume statedump' command, that performs statedumps of
        all the bricks in the volume and saves them in a specified location.
        2. Add new server option 'server.statedump-path'.
        3. Remove multiple function definitions in glusterd.h

Statedump Information:

The 'volume statedump' command performs statedumps on all the bricks in
a given volume. The syntax of the command is,
        gluster volume statedump <VOLNAME> [type]......

Types include,
        * all
        * mem
        * iobuf
        * callpool
        * priv
        * fd
        * inode
Defaults to 'all' when no type is specified.

The statedump files are created by default in /tmp directory of the
server on which the bricks are present.
This path can be changed by setting the 'server.statedump-path' option.

The statedump files will be named as,
        <brick-name>.<pid of brick process>.dump

Change-Id: I01c0e1a8aad490da818e086d89f292bd2ed06fd4
BUG: 1964
Reviewed-on: http://review.gluster.com/321
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Amar Tumballi <amar@gluster.com>

8 files changed, 323 insertions, 26 deletions

diff --git a/xlators/mgmt/glusterd/src/glusterd-handler.c b/xlators/mgmt/glusterd/src/glusterd-handler.c
index 808459cfb7c..4e4b73bf493 100644
--- a/xlators/mgmt/glusterd/src/glusterd-handler.c
+++ b/xlators/mgmt/glusterd/src/glusterd-handler.c
@@ -2784,8 +2784,8 @@ rpcsvc_actor_t gd_svc_cli_actors[] = {
         [GLUSTER_CLI_STATUS_VOLUME]  = {"STATUS_VOLUME", GLUSTER_CLI_STATUS_VOLUME, glusterd_handle_status_volume, NULL, NULL},
         [GLUSTER_CLI_MOUNT]         = { "MOUNT", GLUSTER_CLI_MOUNT, glusterd_handle_mount, NULL, NULL},
         [GLUSTER_CLI_UMOUNT]        = { "UMOUNT", GLUSTER_CLI_UMOUNT, glusterd_handle_umount, NULL, NULL},
-        [GLUSTER_CLI_HEAL_VOLUME]  = { "HEAL_VOLUME", GLUSTER_CLI_HEAL_VOLUME, glusterd_handle_cli_heal_volume, NULL, NULL}
-
+        [GLUSTER_CLI_HEAL_VOLUME]  = { "HEAL_VOLUME", GLUSTER_CLI_HEAL_VOLUME, glusterd_handle_cli_heal_volume, NULL, NULL},
+        [GLUSTER_CLI_STATEDUMP_VOLUME] = {"STATEDUMP_VOLUME", GLUSTER_CLI_STATEDUMP_VOLUME, glusterd_handle_cli_statedump_volume, NULL, NULL},
 };
 
 struct rpcsvc_program gd_svc_cli_prog = {
diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.c b/xlators/mgmt/glusterd/src/glusterd-op-sm.c
index 4271fe2460e..bb8cccfbfeb 100644
--- a/xlators/mgmt/glusterd/src/glusterd-op-sm.c
+++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.c
@@ -1544,6 +1544,7 @@ glusterd_op_build_payload (dict_t **req)
                 case GD_OP_STATUS_VOLUME:
                 case GD_OP_REBALANCE:
                 case GD_OP_HEAL_VOLUME:
+                case GD_OP_STATEDUMP_VOLUME:
                         {
                                 dict_t  *dict = ctx;
                                 dict_copy (dict, req_dict);
@@ -2321,6 +2322,11 @@ glusterd_op_stage_validate (glusterd_op_t op, dict_t *dict, char **op_errstr,
                         ret = glusterd_op_stage_heal_volume (dict, op_errstr);
                         break;
 
+                case GD_OP_STATEDUMP_VOLUME:
+                        ret = glusterd_op_stage_statedump_volume (dict,
+                                                                  op_errstr);
+                        break;
+
                 default:
                         gf_log ("", GF_LOG_ERROR, "Unknown op %d",
                                 op);
@@ -2400,21 +2406,25 @@ glusterd_op_commit_perform (glusterd_op_t op, dict_t *dict, char **op_errstr,
                         ret = glusterd_op_quota (dict, op_errstr);
                         break;
 
-               case GD_OP_LOG_LEVEL:
-                       ret = glusterd_op_log_level (dict);
-                       break;
+                case GD_OP_LOG_LEVEL:
+                        ret = glusterd_op_log_level (dict);
+                        break;
+
+                case GD_OP_STATUS_VOLUME:
+                        ret = glusterd_op_status_volume (dict, op_errstr, rsp_dict);
+                        break;
 
-               case GD_OP_STATUS_VOLUME:
-                       ret = glusterd_op_status_volume (dict, op_errstr, rsp_dict);
-                       break;
+                case GD_OP_REBALANCE:
+                        ret = glusterd_op_rebalance (dict, op_errstr, rsp_dict);
+                        break;
 
-               case GD_OP_REBALANCE:
-                       ret = glusterd_op_rebalance (dict, op_errstr, rsp_dict);
-                       break;
+                case GD_OP_HEAL_VOLUME:
+                        ret = glusterd_op_heal_volume (dict, op_errstr);
+                        break;
 
-               case GD_OP_HEAL_VOLUME:
-                       ret = glusterd_op_heal_volume (dict, op_errstr);
-                       break;
+                case GD_OP_STATEDUMP_VOLUME:
+                        ret = glusterd_op_statedump_volume (dict);
+                        break;
 
                 default:
                         gf_log ("", GF_LOG_ERROR, "Unknown op %d",
@@ -3516,6 +3526,7 @@ glusterd_op_free_ctx (glusterd_op_t op, void *ctx)
                 case GD_OP_STATUS_VOLUME:
                 case GD_OP_REBALANCE:
                 case GD_OP_HEAL_VOLUME:
+                case GD_OP_STATEDUMP_VOLUME:
                         dict_unref (ctx);
                         break;
                 case GD_OP_DELETE_VOLUME:
diff --git a/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c b/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c
index 4d7e687f2e9..615446a6251 100644
--- a/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c
+++ b/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c
@@ -470,6 +470,20 @@ glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret,
                 break;
 
         }
+        case GD_OP_STATEDUMP_VOLUME:
+        {
+                gf1_cli_statedump_vol_rsp rsp = {0,};
+                rsp.op_ret = op_ret;
+                rsp.op_errno = errno;
+                rsp.volname = "";
+                if (op_errstr)
+                        rsp.op_errstr = op_errstr;
+                else
+                        rsp.op_errstr = "";
+                cli_rsp = &rsp;
+                xdrproc = (xdrproc_t) xdr_gf1_cli_statedump_vol_rsp;
+                break;
+        }
         case GD_OP_NONE:
         case GD_OP_MAX:
         {
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index 3681d4b9565..e27d2209f44 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -4161,3 +4161,112 @@ glusterd_is_volume_replicate (glusterd_volinfo_t *volinfo)
                 replicates = _gf_true;
         return replicates;
 }
+
+int
+glusterd_set_dump_options (char *dumpoptions_path, char *options,
+                           int option_cnt)
+{
+        int     ret = -1;
+        char    *dup_options = NULL;
+        char    *option = NULL;
+        char    *tmpptr = NULL;
+        FILE    *fp = NULL;
+
+        if (0 == option_cnt) {
+                ret = 0;
+                goto out;
+        }
+
+        fp = fopen (dumpoptions_path, "w");
+        if (!fp) {
+                ret = -1;
+                goto out;
+        }
+        dup_options = gf_strdup (options);
+        gf_log ("", GF_LOG_INFO, "Recieved following statedump options: %s",
+                dup_options);
+        option = strtok_r (dup_options, " ", &tmpptr);
+        while (option) {
+                fprintf (fp, "%s=yes\n", option);
+                option = strtok_r (NULL, " ", &tmpptr);
+        }
+
+out:
+        if (fp)
+                fclose (fp);
+        return ret;
+}
+
+int
+glusterd_brick_statedump (glusterd_volinfo_t *volinfo,
+                          glusterd_brickinfo_t *brickinfo,
+                          char *options, int option_cnt)
+{
+        int                     ret = -1;
+        xlator_t                *this = NULL;
+        glusterd_conf_t         *conf = NULL;
+        char                    pidfile_path[PATH_MAX] = {0,};
+        char                    path[PATH_MAX] = {0,};
+        char                    dumpoptions_path[PATH_MAX] = {0,};
+        FILE                    *pidfile = NULL;
+        pid_t                   pid = -1;
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        if (uuid_is_null (brickinfo->uuid)) {
+                ret = glusterd_resolve_brick (brickinfo);
+                if (ret) {
+                        gf_log ("glusterd", GF_LOG_ERROR,
+                                "Cannot resolve brick %s:%s",
+                                brickinfo->hostname, brickinfo->path);
+                        goto out;
+                }
+        }
+
+        if (uuid_compare (brickinfo->uuid, conf->uuid)) {
+                ret = 0;
+                goto out;
+        }
+
+        GLUSTERD_GET_VOLUME_DIR (path, volinfo, conf);
+        GLUSTERD_GET_BRICK_PIDFILE (pidfile_path, path, brickinfo->hostname,
+                                    brickinfo->path);
+
+        pidfile = fopen (pidfile_path, "r");
+        if (!pidfile) {
+                gf_log ("", GF_LOG_ERROR, "Unable to open pidfile: %s",
+                        pidfile_path);
+                ret = -1;
+                goto out;
+        }
+
+        ret = fscanf (pidfile, "%d", &pid);
+        if (ret <= 0) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get pid of brick process");
+                ret = -1;
+                goto out;
+        }
+
+        snprintf (dumpoptions_path, sizeof (dumpoptions_path),
+                  "/tmp/glusterdump.%d.options", pid);
+        glusterd_set_dump_options (dumpoptions_path, options, option_cnt);
+
+
+        gf_log ("", GF_LOG_INFO, "Performing statedump on brick with pid %d",
+                pid);
+
+        kill (pid, SIGUSR1);
+
+        sleep (1);
+        unlink (dumpoptions_path);
+
+        ret = 0;
+out:
+        if (pidfile)
+                fclose (pidfile);
+        return ret;
+}
+
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.h b/xlators/mgmt/glusterd/src/glusterd-utils.h
index aca46eae120..3cc137e0579 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.h
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.h
@@ -352,6 +352,11 @@ glusterd_add_brick_to_dict (glusterd_volinfo_t *volinfo,
 gf_boolean_t
 glusterd_is_fuse_available ();
 
+int
+glusterd_brick_statedump (glusterd_volinfo_t *volinfo,
+                          glusterd_brickinfo_t *brickinfo,
+                          char *options, int option_cnt);
+
 gf_boolean_t
 glusterd_is_volume_replicate (glusterd_volinfo_t *volinfo);
 gf_boolean_t
diff --git a/xlators/mgmt/glusterd/src/glusterd-volgen.c b/xlators/mgmt/glusterd/src/glusterd-volgen.c
index e1934493a6b..13c1bffa090 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volgen.c
+++ b/xlators/mgmt/glusterd/src/glusterd-volgen.c
@@ -193,6 +193,7 @@ static struct volopt_map_entry glusterd_volopt_map[] = {
         {VKEY_FEATURES_QUOTA,                    "features/marker",           "quota", "off", NO_DOC, OPT_FLAG_FORCE},
         {VKEY_FEATURES_LIMIT_USAGE,              "features/quota",            "limit-set", NULL, NO_DOC, 0},
         {"features.quota-timeout",               "features/quota",            "timeout", "0", DOC, 0},
+        {"server.statedump-path",                "protocol/server",           "statedump-path", NULL, NO_DOC, 0},
         {NULL,                                                                }
 };
 
diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
index 816ef9b184e..21e7973863a 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
+++ b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
@@ -447,6 +447,65 @@ out:
         return ret;
 }
 
+int
+glusterd_handle_cli_statedump_volume (rpcsvc_request_t *req)
+{
+        int32_t                         ret = -1;
+        gf1_cli_statedump_vol_req       cli_req = {0,};
+        char                            *dup_volname = NULL;
+        char                            *dup_options = NULL;
+        dict_t                          *dict = NULL;
+
+        GF_ASSERT (req);
+
+        ret = -1;
+        if (!xdr_to_generic (req->msg[0], &cli_req,
+                             (xdrproc_t)xdr_gf1_cli_statedump_vol_req)) {
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+        gf_log ("glusterd", GF_LOG_INFO, "Recieved statedump request for "
+                "volume %s with options %s", cli_req.volname, cli_req.options);
+        dict = dict_new ();
+
+        if (!dict)
+                goto out;
+
+        dup_volname = gf_strdup (cli_req.volname);
+        if (!dup_volname)
+                goto out;
+        ret = dict_set_dynstr (dict, "volname", dup_volname);
+        if (ret)
+                goto out;
+
+        dup_options = gf_strdup(cli_req.options);
+        if (!dup_volname)
+                goto out;
+        ret = dict_set_dynstr (dict, "options", dup_options);
+        if (ret)
+                goto out;
+
+        ret = dict_set_int32 (dict, "option_cnt", cli_req.option_cnt);
+        if (ret)
+                goto out;
+
+        ret = glusterd_op_begin (req, GD_OP_STATEDUMP_VOLUME, dict);
+
+        gf_cmd_log ("statedump", "on volume %s %s", cli_req.volname,
+                    ((0 == ret) ? "SUCCEEDED" : "FAILED"));
+
+out:
+        if (ret && dict)
+                dict_unref (dict);
+        if (cli_req.volname)
+                free (cli_req.volname);
+        if (cli_req.options)
+                free (cli_req.options);
+        glusterd_friend_sm ();
+        glusterd_op_sm();
+
+        return ret;
+}
 
 /* op-sm */
 int
@@ -610,6 +669,37 @@ out:
 }
 
 int
+glusterd_op_statedump_volume_args_get (dict_t *dict, char **volname,
+                                       char **options, int *option_cnt)
+{
+        int ret = -1;
+
+        if (!dict || !volname || !options || !option_cnt)
+                goto out;
+
+        ret = dict_get_str (dict, "volname", volname);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get volname");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "options", options);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get options");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "option_cnt", option_cnt);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get option count");
+                goto out;
+        }
+
+out:
+        return ret;
+}
+
+int
 glusterd_op_stage_start_volume (dict_t *dict, char **op_errstr)
 {
         int                                     ret = 0;
@@ -905,6 +995,46 @@ out:
 }
 
 int
+glusterd_op_stage_statedump_volume (dict_t *dict, char **op_errstr)
+{
+        int                     ret = -1;
+        char                    *volname = NULL;
+        char                    *options = NULL;
+        int                     option_cnt = 0;
+        gf_boolean_t            is_running = _gf_false;
+        glusterd_volinfo_t      *volinfo = NULL;
+        char                    msg[2408] = {0,};
+
+        ret = glusterd_op_statedump_volume_args_get (dict, &volname, &options,
+                                                     &option_cnt);
+        if (ret)
+                goto out;
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                snprintf (msg, sizeof(msg), "Volume %s does not exist",
+                          volname);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        is_running = glusterd_is_volume_started (volinfo);
+        if (!is_running) {
+                snprintf (msg, sizeof(msg), "Volume %s is not in a started"
+                          " state", volname);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                ret = -1;
+                goto out;
+        }
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
 glusterd_op_create_volume (dict_t *dict, char **op_errstr)
 {
         int                   ret        = 0;
@@ -1204,3 +1334,34 @@ glusterd_op_heal_volume (dict_t *dict, char **op_errstr)
 
         return ret;
 }
+
+int
+glusterd_op_statedump_volume (dict_t *dict)
+{
+        int                     ret = 0;
+        char                    *volname = NULL;
+        char                    *options = NULL;
+        int                     option_cnt = 0;
+        glusterd_volinfo_t      *volinfo = NULL;
+        glusterd_brickinfo_t    *brickinfo = NULL;
+
+        ret = glusterd_op_statedump_volume_args_get (dict, &volname, &options,
+                                                     &option_cnt);
+        if (ret)
+                goto out;
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret)
+                goto out;
+        gf_log ("", GF_LOG_DEBUG, "Performing statedump on volume %s", volname);
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                ret = glusterd_brick_statedump (volinfo, brickinfo, options,
+                                                option_cnt);
+                if (ret)
+                        goto out;
+        }
+
+out:
+        return ret;
+}
+
diff --git a/xlators/mgmt/glusterd/src/glusterd.h b/xlators/mgmt/glusterd/src/glusterd.h
index f5a10189b0e..203f6e975c7 100644
--- a/xlators/mgmt/glusterd/src/glusterd.h
+++ b/xlators/mgmt/glusterd/src/glusterd.h
@@ -80,6 +80,7 @@ typedef enum glusterd_op_ {
         GD_OP_STATUS_VOLUME,
         GD_OP_REBALANCE,
         GD_OP_HEAL_VOLUME,
+        GD_OP_STATEDUMP_VOLUME,
         GD_OP_MAX,
 } glusterd_op_t;
 
@@ -541,17 +542,9 @@ glusterd_handle_log_level (rpcsvc_request_t *req);
 /* handler functions */
 int32_t glusterd_op_begin (rpcsvc_request_t *req, glusterd_op_t op, void *ctx);
 
-int glusterd_handle_gsync_set (rpcsvc_request_t *req);
-int glusterd_handle_quota (rpcsvc_request_t *req);
-int glusterd_handle_replace_brick (rpcsvc_request_t *req);
-int glusterd_handle_log_filename (rpcsvc_request_t *req);
-int glusterd_handle_log_locate (rpcsvc_request_t *req);
-int glusterd_handle_log_level (rpcsvc_request_t *req);
-int glusterd_handle_log_rotate (rpcsvc_request_t *req);
-int glusterd_handle_create_volume (rpcsvc_request_t *req);
-int glusterd_handle_cli_start_volume (rpcsvc_request_t *req);
-int glusterd_handle_cli_stop_volume (rpcsvc_request_t *req);
-int glusterd_handle_cli_delete_volume (rpcsvc_request_t *req);
+/* removed other definitions as they have been defined elsewhere in this file*/
+
+int glusterd_handle_cli_statedump_volume (rpcsvc_request_t *req);
 
 int glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr,
                                   size_t len, int cmd, defrag_cbk_fn_t cbk);
@@ -590,12 +583,15 @@ int glusterd_op_stage_remove_brick (dict_t *dict, char **op_errstr);
 int glusterd_op_stage_rebalance (dict_t *dict, char **op_errstr);
 int glusterd_op_rebalance (dict_t *dict, char **op_errstr, dict_t *rsp_dict);
 
+int glusterd_op_stage_statedump_volume (dict_t *dict, char **op_errstr);
+int glusterd_op_statedump_volume (dict_t *dict);
 
 /* misc */
 void glusterd_do_replace_brick (void *data);
 int glusterd_op_perform_remove_brick (glusterd_volinfo_t  *volinfo, char *brick,
                                       int force, int *need_migrate);
 int glusterd_op_stop_volume_args_get (dict_t *dict, char** volname, int *flags);
-
+int glusterd_op_statedump_volume_args_get (dict_t *dict, char **volname,
+                                           char **options, int *option_cnt);
 
 #endif