glusterd: Extended glusterd_is_service_running to get svcs's pid.

Also, volume status cmd would print "N/A" if pid couldn't be
retrieved from pidfile.

Change-Id: Ie83d228b1cf86397d181885b325e337a403e6ed2
BUG: 3043
Reviewed-on: http://review.gluster.com/650
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Amar Tumballi <amar@gluster.com>
Reviewed-by: Vijay Bellur <vijay@gluster.com>

5 files changed, 35 insertions, 51 deletions

diff --git a/cli/src/cli-cmd-volume.c b/cli/src/cli-cmd-volume.c
index 18d17b4546d..b5a177a869d 100644
--- a/cli/src/cli-cmd-volume.c
+++ b/cli/src/cli-cmd-volume.c
@@ -1376,7 +1376,7 @@ cli_cmd_volume_status_cbk (struct cli_state *state,
 
 
 int
-cli_print_brick_status (char *brick, int port, int online, int pid)
+cli_print_brick_status (char *brick, int port, int online, char *pid)
 {
         int  fieldlen = CLI_VOL_STATUS_BRICK_LEN;
         char buf[80] = {0,};
@@ -1400,7 +1400,7 @@ cli_print_brick_status (char *brick, int port, int online, int pid)
                         printf ("%s", p);
                         while (num_tabs-- != 0)
                                 printf ("\t");
-                        cli_out ("%d\t%c\t%d", port, online?'Y':'N', pid);
+                        cli_out ("%d\t%c\t%s", port, online?'Y':'N', pid);
                         bricklen = 0;
                 }
         }
diff --git a/cli/src/cli-rpc-ops.c b/cli/src/cli-rpc-ops.c
index 2841a32d589..8a88e57d595 100644
--- a/cli/src/cli-rpc-ops.c
+++ b/cli/src/cli-rpc-ops.c
@@ -3807,7 +3807,8 @@ gf_cli3_1_status_cbk (struct rpc_req *req, struct iovec *iov,
         int                             port = 0;
         int                             online = 0;
         char                            key[1024] = {0,};
-        int                             pid = 0;
+        int                             pid = -1;
+        char                            *pid_str = NULL;
         char                            brick[8192] = {0,};
         char                            *volname = NULL;
 
@@ -3875,11 +3876,22 @@ gf_cli3_1_status_cbk (struct rpc_req *req, struct iovec *iov,
                 memset (key, 0, sizeof (key));
                 snprintf (key, sizeof (key), "brick%d.pid", i);
                 ret = dict_get_int32 (dict, key, &pid);
+                if (ret)
+                        goto out;
+                if (pid == -1)
+                        ret = gf_asprintf (&pid_str, "%s", "N/A");
+                else
+                        ret = gf_asprintf (&pid_str, "%d", pid);
+
+                if (ret == -1)
+                        goto out;
 
                 snprintf (brick, sizeof (brick) -1, "%s:%s", hostname, path);
 
                 cli_print_line (CLI_BRICK_STATUS_LINE_LEN);
-                cli_print_brick_status (brick, port, online, pid);
+                cli_print_brick_status (brick, port, online, pid_str);
+                if (pid_str)
+                        GF_FREE (pid_str);
         }
 
         ret = rsp.op_ret;
diff --git a/cli/src/cli.h b/cli/src/cli.h
index 70310f103ed..1d2e069735a 100644
--- a/cli/src/cli.h
+++ b/cli/src/cli.h
@@ -236,7 +236,7 @@ cli_cmd_volume_status_parse (const char **words, int wordcount,
                              dict_t **options);
 
 int
-cli_print_brick_status (char *brick, int port, int online, int pid);
+cli_print_brick_status (char *brick, int port, int online, char *pid);
 
 void
 cli_print_line (int len);
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index 1e797a88200..2eaf4f4a2f4 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -2308,12 +2308,13 @@ out:
         return ret;
 }
 
+/* Valid only in if service is 'local' to glusterd.
+ * pid can be -1, if reading pidfile failed */
 gf_boolean_t
-glusterd_is_service_running (char *pidfile)
+glusterd_is_service_running (char *pidfile, int *pid)
 {
         FILE            *file = NULL;
         gf_boolean_t    running = _gf_false;
-        gf_boolean_t    locked = _gf_false;
         int             ret = 0;
         int             fno = 0;
 
@@ -2322,20 +2323,20 @@ glusterd_is_service_running (char *pidfile)
                 goto out;
 
         fno = fileno (file);
-        ret = lockf (fno, F_TLOCK, 0);
-        if (!ret) {
-                locked = _gf_true;
+        ret = lockf (fno, F_TEST, 0);
+        if (ret == -1)
+                running = _gf_true;
+        if (!pid)
                 goto out;
+
+        ret = fscanf (file, "%d", pid);
+        if (ret <= 0) {
+                gf_log ("", GF_LOG_ERROR, "Unable to read pidfile: %s, %s",
+                        pidfile, strerror (errno));
+                *pid = -1;
         }
 
-        running = _gf_true;
 out:
-        if (locked) {
-                GF_ASSERT (file);
-                if (lockf (fno, F_ULOCK, 0) < 0)
-                        gf_log ("", GF_LOG_WARNING, "Cannot unlock pidfile: %s"
-                                " reason: %s", pidfile, strerror(errno));
-        }
         if (file)
                 fclose (file);
         return running;
@@ -2586,7 +2587,7 @@ glusterd_is_nodesvc_running (char *server)
 
         glusterd_get_nodesvc_pidfile (server, priv->workdir,
                                             pidfile, sizeof (pidfile));
-        return glusterd_is_service_running (pidfile);
+        return glusterd_is_service_running (pidfile, NULL);
 }
 
 int32_t
@@ -3094,8 +3095,6 @@ glusterd_add_brick_to_dict (glusterd_volinfo_t *volinfo,
         char            base_key[8192] = {0};
         char            pidfile[PATH_MAX] = {0};
         char            path[PATH_MAX] = {0};
-        FILE            *file = NULL;
-        int             fd = -1;
         int32_t         pid = -1;
         int32_t         brick_online = -1;
         xlator_t        *this = NULL;
@@ -3128,39 +3127,12 @@ glusterd_add_brick_to_dict (glusterd_volinfo_t *volinfo,
         if (ret)
                 goto out;
 
-
         GLUSTERD_GET_VOLUME_DIR (path, volinfo, priv);
         GLUSTERD_GET_BRICK_PIDFILE (pidfile, path, brickinfo->hostname,
                                     brickinfo->path);
 
-        file = fopen (pidfile, "r");
-        if (!file) {
-                gf_log ("", GF_LOG_ERROR, "Unable to open pidfile: %s",
-                        pidfile);
-                /* pidfile doesn't exist means brick is down*/
-                pid = 0;
-                brick_online = 0;
-                goto cont;
-        } else {
-                ret = fscanf (file, "%d", &pid);
-                if (ret <= 0) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to read pidfile: %s",
-                                pidfile);
-                        ret = -1;
-                        goto out;
-                }
-
-                /* check if process is crashed*/
-                fd = fileno (file);
-                if ((fd != -1) && (lockf (fd, F_TEST, 0)))
-                        brick_online = 1;
-                else {
-                        pid = 0;
-                        brick_online = 0;
-                }
-        }
+        brick_online = glusterd_is_service_running (pidfile, &pid);
 
-cont:
         memset (key, 0, sizeof (key));
         snprintf (key, sizeof (key), "%s.pid", base_key);
         ret = dict_set_int32 (dict, key, pid);
@@ -3174,9 +3146,6 @@ cont:
                 goto out;
 
 out:
-        if (file)
-                fclose (file);
-
         if (ret)
                 gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
 
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.h b/xlators/mgmt/glusterd/src/glusterd-utils.h
index 8401b61bc16..a34956b0c4a 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.h
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.h
@@ -154,6 +154,9 @@ glusterd_get_nodesvc_volfile (char *server, char *workdir,
                                     char *volfile, size_t len);
 
 gf_boolean_t
+glusterd_is_service_running (char *pidfile, int *pid);
+
+gf_boolean_t
 glusterd_is_nodesvc_running ();
 
 gf_boolean_t