diff --git a/lustre/include/obd_support.h b/lustre/include/obd_support.h index 6dcb0d90d8527c2feedd201f89ded4d724c6af8f..e97b58d908a8c0d6390d0088fabf1409f211bbf1 100644 --- a/lustre/include/obd_support.h +++ b/lustre/include/obd_support.h @@ -60,7 +60,7 @@ extern unsigned int obd_alloc_fail_rate; #endif /* Time to wait for all clients to reconnect during recovery */ /* Should be very conservative; must catch the first reconnect after reboot */ -#define OBD_RECOVERY_FACTOR (5 / 2) /* times obd_timeout */ +#define OBD_RECOVERY_FACTOR (3) /* times obd_timeout */ /* Change recovery-small 26b time if you change this */ #define PING_INTERVAL max(obd_timeout / 4, 1U) /* Client may skip 1 ping; we must wait at least 2.5. But for multiple diff --git a/lustre/ldlm/ldlm_lib.c b/lustre/ldlm/ldlm_lib.c index e134fbe8c699f0248a08473c83d4cb1ce23d5167..b17366e402d2cae176dcd6225a602328c78b9965 100644 --- a/lustre/ldlm/ldlm_lib.c +++ b/lustre/ldlm/ldlm_lib.c @@ -1282,7 +1282,8 @@ static void process_recovery_queue(struct obd_device *obd) DEBUG_REQ(D_HA, req, "processing: "); (void)obd->obd_recovery_handler(req); obd->obd_replayed_requests++; - reset_recovery_timer(obd, 2 * /* safety */ + reset_recovery_timer(obd, OBD_RECOVERY_FACTOR * + AT_OFF ? obd_timeout : at_get(&req->rq_rqbd->rqbd_service->srv_at_estimate), 1); /* bug 1580: decide how to properly sync() in recovery */ //mds_fsync_super(obd->u.obt.obt_sb); diff --git a/lustre/ptlrpc/pinger.c b/lustre/ptlrpc/pinger.c index 32bd0ea9f4302014fa171fa516c3ab2dca63fdea..59ccff3ba27d95438e0f8b40dbbc350d6010299e 100644 --- a/lustre/ptlrpc/pinger.c +++ b/lustre/ptlrpc/pinger.c @@ -70,6 +70,7 @@ void ptlrpc_update_next_ping(struct obd_import *imp) int time = PING_INTERVAL; if (imp->imp_state == LUSTRE_IMP_DISCON) { int dtime = max_t(int, CONNECTION_SWITCH_MIN, + AT_OFF ? 0 : at_get(&imp->imp_at.iat_net_latency)); time = min(time, dtime); }