glusterd: persist brickinfo->real_path

Since real_path was not persisted and gets constructed at every glusterd
restart, glusterd will fail to come up if one of the brick's underlying file
system is crashed.

Solution is to construct real_path only once and get it persisted.

Change-Id: I97abc30372c1ffbbb2d43b716d7af09172147b47
BUG: 1330481
Signed-off-by: Atin Mukherjee <amukherj@redhat.com>
Reviewed-on: http://review.gluster.org/14075
CentOS-regression: Gluster Build System <jenkins@build.gluster.com>
Smoke: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Kaushal M <kaushal@redhat.com>
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>

3 files changed, 39 insertions, 25 deletions

diff --git a/xlators/mgmt/glusterd/src/glusterd-store.c b/xlators/mgmt/glusterd/src/glusterd-store.c
index 605c4c2519f..aa1201e492d 100644
--- a/xlators/mgmt/glusterd/src/glusterd-store.c
+++ b/xlators/mgmt/glusterd/src/glusterd-store.c
@@ -379,6 +379,11 @@ glusterd_store_brickinfo_write (int fd, glusterd_brickinfo_t *brickinfo)
         if (ret)
                 goto out;
 
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_REAL_PATH,
+                                   brickinfo->path);
+        if (ret)
+                goto out;
+
         snprintf (value, sizeof(value), "%d", brickinfo->port);
         ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_PORT, value);
 
@@ -2305,6 +2310,10 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo)
                                     strlen (GLUSTERD_STORE_KEY_BRICK_PATH))) {
                                 strncpy (brickinfo->path, value,
                                          sizeof (brickinfo->path));
+                        } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_REAL_PATH,
+                                    strlen (GLUSTERD_STORE_KEY_BRICK_REAL_PATH))) {
+                                strncpy (brickinfo->real_path, value,
+                                         sizeof (brickinfo->real_path));
                         } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_PORT,
                                     strlen (GLUSTERD_STORE_KEY_BRICK_PORT))) {
                                 gf_string2int (value, &brickinfo->port);
@@ -2405,8 +2414,8 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo)
                  * snapshot or snapshot restored volume this would be done post
                  * creating the brick mounts
                  */
-                if (!volinfo->is_snap_volume &&
-                    gf_uuid_is_null (volinfo->restored_from_snap)) {
+                if (brickinfo->real_path[0] == '\0' && !volinfo->is_snap_volume
+                    && gf_uuid_is_null (volinfo->restored_from_snap)) {
                         /* By now if the brick is a local brick then it will be
                          * able to resolve which is the only thing we want now
                          * for checking  whether the brickinfo->uuid matches
@@ -3300,19 +3309,21 @@ glusterd_recreate_vol_brick_mounts (xlator_t  *this,
                                 "Failed to mount brick_mount_path");
                 }
                 if (!gf_uuid_compare(brickinfo->uuid, MY_UUID)) {
-                        if (!realpath (brickinfo->path, abspath)) {
-                                gf_msg (this->name, GF_LOG_CRITICAL,
-                                        errno,
-                                        GD_MSG_BRICKINFO_CREATE_FAIL,
-                                        "realpath() failed for brick %s"
-                                        ". The underlying file system "
-                                        "may be in bad state",
-                                        brickinfo->path);
-                                ret = -1;
-                                goto out;
+                        if (brickinfo->real_path[0] == '\0') {
+                                if (!realpath (brickinfo->path, abspath)) {
+                                        gf_msg (this->name, GF_LOG_CRITICAL,
+                                                errno,
+                                                GD_MSG_BRICKINFO_CREATE_FAIL,
+                                                "realpath() failed for brick %s"
+                                                ". The underlying file system "
+                                                "may be in bad state",
+                                                brickinfo->path);
+                                        ret = -1;
+                                        goto out;
+                                }
+                                strncpy (brickinfo->real_path, abspath,
+                                         strlen(abspath));
                         }
-                        strncpy (brickinfo->real_path, abspath,
-                                 strlen(abspath));
                 }
 
                 if (brick_mount_path) {
diff --git a/xlators/mgmt/glusterd/src/glusterd-store.h b/xlators/mgmt/glusterd/src/glusterd-store.h
index 0f4cc9b138b..ec293b64e82 100644
--- a/xlators/mgmt/glusterd/src/glusterd-store.h
+++ b/xlators/mgmt/glusterd/src/glusterd-store.h
@@ -85,6 +85,7 @@ typedef enum glusterd_store_ver_ac_{
 
 #define GLUSTERD_STORE_KEY_BRICK_HOSTNAME       "hostname"
 #define GLUSTERD_STORE_KEY_BRICK_PATH           "path"
+#define GLUSTERD_STORE_KEY_BRICK_REAL_PATH      "real_path"
 #define GLUSTERD_STORE_KEY_BRICK_PORT           "listen-port"
 #define GLUSTERD_STORE_KEY_BRICK_RDMA_PORT      "rdma.listen-port"
 #define GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED "decommissioned"
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index 216089db85a..68208312acb 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -1109,7 +1109,7 @@ glusterd_brickinfo_new_from_brick (char *brick,
         strncpy (new_brickinfo->hostname, hostname, 1024);
         strncpy (new_brickinfo->path, path, 1024);
 
-        if (construct_real_path) {
+        if (construct_real_path && new_brickinfo->real_path[0] == '\0') {
                 if (!realpath (new_brickinfo->path, abspath)) {
                         /* ENOENT indicates that brick path has not been created
                          * which is a valid scenario */
@@ -3169,17 +3169,19 @@ glusterd_import_new_brick (dict_t *peer_data, int32_t vol_count,
 
         gf_uuid_parse (brick_uuid_str, new_brickinfo->uuid);
         if (!gf_uuid_compare(new_brickinfo->uuid, MY_UUID)) {
-                if (!realpath (new_brickinfo->path, abspath)) {
-                        gf_msg (this->name, GF_LOG_CRITICAL, errno,
-                                GD_MSG_BRICKINFO_CREATE_FAIL, "realpath() "
-                                "failed for brick %s. The underlying file "
-                                "system may be in bad state",
-                                new_brickinfo->path);
-                        ret = -1;
-                        goto out;
+                if (new_brickinfo->real_path[0] == '\0') {
+                        if (!realpath (new_brickinfo->path, abspath)) {
+                                gf_msg (this->name, GF_LOG_CRITICAL, errno,
+                                        GD_MSG_BRICKINFO_CREATE_FAIL,
+                                        "realpath() failed for brick %s. The "
+                                        "underlying file system may be in bad "
+                                        "state", new_brickinfo->path);
+                                ret = -1;
+                                goto out;
+                        }
+                        strncpy (new_brickinfo->real_path, abspath,
+                                 strlen(abspath));
                 }
-                strncpy (new_brickinfo->real_path, abspath,
-                         strlen(abspath));
         }
 
         *brickinfo = new_brickinfo;