From: Alexander Boyko <alexander_boyko@xxxxxxxxxxx> Once connected, the previously gathered AT statistics is not valid anymore because may reflect other routing, etc. The connect by itself could take a long time due to different reasons (e.g. server was not ready) and net latency got very high (see import_select_connection()) what does not reflect the current situation. Take into account only the current (re-)CONNECT rpc latency. Signed-off-by: Vitaly Fertman <vitaly_fertman@xxxxxxxxxxx> Signed-off-by: Alexander Boyko <alexander_boyko@xxxxxxxxxxx> Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5380 Xyratex-bug-id: MRP-1285 Reviewed-on: http://review.whamcloud.com/11155 Reviewed-by: Liang Zhen <liang.zhen@xxxxxxxxx> Reviewed-by: Andreas Dilger <andreas.dilger@xxxxxxxxx> Reviewed-by: Mike Pershin <mike.pershin@xxxxxxxxx> Signed-off-by: James Simmons <jsimmons@xxxxxxxxxxxxx> --- drivers/staging/lustre/lustre/ptlrpc/client.c | 4 ++-- drivers/staging/lustre/lustre/ptlrpc/import.c | 8 ++++++++ .../staging/lustre/lustre/ptlrpc/ptlrpc_internal.h | 2 ++ 3 files changed, 12 insertions(+), 2 deletions(-) diff --git a/drivers/staging/lustre/lustre/ptlrpc/client.c b/drivers/staging/lustre/lustre/ptlrpc/client.c index bea1c16..27fe00d 100644 --- a/drivers/staging/lustre/lustre/ptlrpc/client.c +++ b/drivers/staging/lustre/lustre/ptlrpc/client.c @@ -283,8 +283,8 @@ int ptlrpc_at_get_net_latency(struct ptlrpc_request *req) } /* Adjust expected network latency */ -static void ptlrpc_at_adj_net_latency(struct ptlrpc_request *req, - unsigned int service_time) +void ptlrpc_at_adj_net_latency(struct ptlrpc_request *req, + unsigned int service_time) { unsigned int nl, oldnl; struct imp_at *at; diff --git a/drivers/staging/lustre/lustre/ptlrpc/import.c b/drivers/staging/lustre/lustre/ptlrpc/import.c index 013a957..17f3fec 100644 --- a/drivers/staging/lustre/lustre/ptlrpc/import.c +++ b/drivers/staging/lustre/lustre/ptlrpc/import.c @@ -846,6 +846,14 @@ static int ptlrpc_connect_interpret(const struct lu_env *env, imp->imp_obd->obd_self_export->exp_connect_data = *ocd; class_export_put(exp); + /* + * The net statistics after (re-)connect is not valid anymore, + * because may reflect other routing, etc. + */ + at_init(&imp->imp_at.iat_net_latency, 0, 0); + ptlrpc_at_adj_net_latency(request, + lustre_msg_get_service_time(request->rq_repmsg)); + obd_import_event(imp->imp_obd, imp, IMP_EVENT_OCD); if (aa->pcaa_initial_connect) { diff --git a/drivers/staging/lustre/lustre/ptlrpc/ptlrpc_internal.h b/drivers/staging/lustre/lustre/ptlrpc/ptlrpc_internal.h index a9831fa..29cfac2 100644 --- a/drivers/staging/lustre/lustre/ptlrpc/ptlrpc_internal.h +++ b/drivers/staging/lustre/lustre/ptlrpc/ptlrpc_internal.h @@ -53,6 +53,8 @@ int ptlrpc_start_thread(struct ptlrpc_service_part *svcpt, int wait); int ptlrpcd_start(struct ptlrpcd_ctl *pc); /* client.c */ +void ptlrpc_at_adj_net_latency(struct ptlrpc_request *req, + unsigned int service_time); struct ptlrpc_bulk_desc *ptlrpc_new_bulk(unsigned npages, unsigned max_brw, unsigned type, unsigned portal); int ptlrpc_request_cache_init(void); -- 1.7.1 _______________________________________________ devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxx http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
