LU-10045 obdclass: multiple try when register target

It is possible that the connection between MGC and MGS has not
been established when register target to MGS for server mount.
At that time, the ptlrpcd may be trying to (re-)connect to MGS
at background. Under such case, the mount process should not
report failure (-ESHUTDOWN -r -EIO), instead, it can retry the
MGS_TARGET_REG RPC after sometime (such as 2 seconds).

Signed-off-by: Fan Yong <fan.yong@intel.com>
Change-Id: I44e53a9d1de037907bdb5148b8c44d332439a50c
Reviewed-on: https://review.whamcloud.com/30761
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Reviewed-by: Alex Zhuravlev <alexey.zhuravlev@intel.com>
Reviewed-by: Faccini Bruno <bruno.faccini@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

diff --git a/lustre/obdclass/obd_mount_server.c b/lustre/obdclass/obd_mount_server.c
index cbe627e..8441d5b 100644 (file)
--- a/lustre/obdclass/obd_mount_server.c
+++ b/lustre/obdclass/obd_mount_server.c
@@ -1187,6 +1187,7 @@ static int server_register_target(struct lustre_sb_info *lsi)
        struct mgs_target_info *mti = NULL;
        bool writeconf;
        int rc;
+       int tried = 0;
        ENTRY;
 
        LASSERT(mgc);
@@ -1211,6 +1212,7 @@ static int server_register_target(struct lustre_sb_info *lsi)
        writeconf = !!(lsi->lsi_flags & (LDD_F_NEED_INDEX | LDD_F_UPDATE));
        mti->mti_flags |= LDD_F_OPC_REG;
 
+again:
        /* Register the target */
        /* FIXME use mgc_process_config instead */
        rc = obd_set_info_async(NULL, mgc->u.cli.cl_mgc_mgsexp,
@@ -1224,6 +1226,17 @@ static int server_register_target(struct lustre_sb_info *lsi)
                                "to start: rc = %d. Please see messages on "
                                "the MGS.\n", lsi->lsi_svname, rc);
                } else if (writeconf) {
+                       if ((rc == -ESHUTDOWN || rc == -EIO) && ++tried < 5) {
+                               /* The connection with MGS is not established.
+                                * Try again after 2 seconds. Interruptable. */
+                               set_current_state(TASK_INTERRUPTIBLE);
+                               schedule_timeout(
+                                       msecs_to_jiffies(MSEC_PER_SEC) * 2);
+                               set_current_state(TASK_RUNNING);
+                               if (!signal_pending(current))
+                                       goto again;
+                       }
+
                        LCONSOLE_ERROR_MSG(0x15f,
                                "%s: cannot register this server with the MGS: "
                                "rc = %d. Is the MGS running?\n",