argv[0], argv[1], argv[2], rc);
} else {
- CERROR("Invoked upcall %s %s %s\n",
- argv[0], argv[1], argv[2]);
+ CWARN("Invoked upcall %s %s %s\n",
+ argv[0], argv[1], argv[2]);
}
}
rc = USERMODEHELPER(argv[0], argv, envp);
if (rc < 0) {
CERROR("Error invoking recovery upcall %s %s %s %s %s: %d; "
- "check /proc/sys/lustre/lustre_upcall\n",
+ "check /proc/sys/lustre/upcall\n",
argv[0], argv[1], argv[2], argv[3], argv[4],rc);
} else {
- CERROR("Invoked upcall %s %s %s %s %s\n",
- argv[0], argv[1], argv[2], argv[3], argv[4]);
+ CWARN("Invoked upcall %s %s %s %s %s\n",
+ argv[0], argv[1], argv[2], argv[3], argv[4]);
}
#else
if (imp->imp_state == LUSTRE_IMP_CLOSED) {
LASSERT (obd_lustre_upcall != NULL);
if (strcmp(obd_lustre_upcall, "DEFAULT") == 0) {
- CDEBUG(D_ERROR, "%s: starting recovery without upcall\n",
+ CDEBUG(D_HA, "%s: starting recovery without upcall\n",
imp->imp_target_uuid.uuid);
ptlrpc_connect_import(imp, NULL);
}
else if (strcmp(obd_lustre_upcall, "NONE") == 0) {
- CDEBUG(D_ERROR, "%s: recovery diabled\n",
+ CDEBUG(D_HA, "%s: recovery disabled\n",
imp->imp_target_uuid.uuid);
}
else {
- CDEBUG(D_ERROR, "%s: calling upcall to start recovery\n",
+ CDEBUG(D_HA, "%s: calling upcall to start recovery\n",
imp->imp_target_uuid.uuid);
ptlrpc_run_failed_import_upcall(imp);
}
{
int rc = 0;
struct list_head *tmp, *pos;
- struct ptlrpc_request *req;
+ struct ptlrpc_request *req = NULL;
unsigned long flags;
__u64 last_transno;
ENTRY;
*/
list_for_each_safe(tmp, pos, &imp->imp_replay_list) {
req = list_entry(tmp, struct ptlrpc_request, rq_replay_list);
+
+ /* If need to resend, stop on the matching one first. It's
+ possible though it's already been committed, so in that case
+ we'll just continue with replay */
+ if (imp->imp_resend_replay &&
+ req->rq_transno == last_transno) {
+ lustre_msg_add_flags(req->rq_reqmsg, MSG_RESENT);
+ break;
+ }
+
if (req->rq_transno > last_transno) {
- rc = ptlrpc_replay_req(req);
- if (rc) {
- CERROR("recovery replay error %d for req "
- LPD64"\n", rc, req->rq_xid);
- RETURN(rc);
- }
- *inflight = 1;
+ imp->imp_last_replay_transno = req->rq_transno;
break;
}
+
+ req = NULL;
+ }
+
+ imp->imp_resend_replay = 0;
+
+ if (req != NULL) {
+ rc = ptlrpc_replay_req(req);
+ if (rc) {
+ CERROR("recovery replay error %d for req "
+ LPD64"\n", rc, req->rq_xid);
+ RETURN(rc);
+ }
+ *inflight = 1;
}
RETURN(rc);
}
rc = ptlrpc_connect_import(imp, NULL);
}
-
/* Wait for recovery to complete and resend. If evicted, then
this request will be errored out later.*/
spin_lock_irqsave(&failed_req->rq_lock, flags);
failed_req->rq_resend = 1;
spin_unlock_irqrestore(&failed_req->rq_lock, flags);
-
+
EXIT;
}
* requests. */
if (!active) {
ptlrpc_invalidate_import(imp, 0);
- }
+ imp->imp_deactive = 1;
+ }
/* When activating, mark import valid, and attempt recovery */
if (active) {
+ imp->imp_deactive = 0;
CDEBUG(D_HA, "setting import %s VALID\n",
imp->imp_target_uuid.uuid);
rc = ptlrpc_recover_import(imp, NULL);
{
int rc;
ENTRY;
-
+
/* force import to be disconnected. */
ptlrpc_set_import_discon(imp);
-
+
rc = ptlrpc_recover_import_no_retry(imp, new_uuid);
RETURN(rc);
if (rc)
RETURN(rc);
- CDEBUG(D_ERROR, "%s: recovery started, waiting\n",
+ CDEBUG(D_HA, "%s: recovery started, waiting\n",
imp->imp_target_uuid.uuid);
lwi = LWI_TIMEOUT(MAX(obd_timeout * HZ, 1), NULL, NULL);
rc = l_wait_event(imp->imp_recovery_waitq,
!ptlrpc_import_in_recovery(imp), &lwi);
- CDEBUG(D_ERROR, "%s: recovery finished\n",
+ CDEBUG(D_HA, "%s: recovery finished\n",
imp->imp_target_uuid.uuid);
RETURN(rc);
CDEBUG(D_HA, "disconnecting export %p/%s\n",
exp, exp->exp_client_uuid.uuid);
+ if (obd_dump_on_timeout)
+ portals_debug_dumplog();
+
/* Most callers into obd_disconnect are removing their own reference
* (request, for example) in addition to the one from the hash table.
* We don't have such a reference here, so make one. */