DAOS-16469 object: debug patch for EC16P2GX performance - b26

daos-stack · Sep 9, 2024 · df1b543 · df1b543
1 parent 70e4362
commit df1b543
Show file tree

Hide file tree

Showing 4 changed files with 32 additions and 7 deletions.
diff --git a/src/dtx/dtx_common.c b/src/dtx/dtx_common.c
@@ -755,7 +755,7 @@ dtx_batched_commit(void *arg)
 			dbca->dbca_cleanup_done = 0;
 		}
 
-		if (dtx_cont_opened(cont) &&
+		if (0 && dtx_cont_opened(cont) &&
 		    !dbca->dbca_deregister && dbca->dbca_cleanup_req == NULL &&
 		    stat.dtx_oldest_active_time != 0 &&
 		    dtx_sec2age(stat.dtx_oldest_active_time) >= dbca->dbca_cleanup_thd) {
@@ -1351,7 +1351,7 @@ dtx_leader_end(struct dtx_leader_handle *dlh, struct ds_cont_hdl *coh, int resul
 		}
 	}
 
-	if (dth->dth_prepared || dtx_batched_ult_max == 0) {
+	if (dtx_batched_ult_max == 0) {
 		dth->dth_sync = 1;
 		goto sync;
 	}

diff --git a/src/dtx/dtx_rpc.c b/src/dtx/dtx_rpc.c
@@ -225,9 +225,10 @@ dtx_req_cb(const struct crt_cb_info *cb_info)
 	}
 
 out:
-	D_DEBUG(DB_TRACE, "DTX req for opc %x (req %p future %p) got reply from %d/%d: "
-		"epoch :"DF_X64", result %d\n", dra->dra_opc, req, dra->dra_future,
-		drr->drr_rank, drr->drr_tag, din != NULL ? din->di_epoch : 0, rc);
+	DL_CDEBUG(rc < 0 && rc != -DER_NONEXIST, DLOG_ERR, DB_TRACE, rc,
+		  "DTX req for opc %x (req %p future %p) got reply from %d/%d: "
+		  "epoch :"DF_X64"\n", dra->dra_opc, req, dra->dra_future,
+		  drr->drr_rank, drr->drr_tag, din != NULL ? din->di_epoch : 0);
 
 	drr->drr_comp = 1;
 	drr->drr_result = rc;

diff --git a/src/object/cli_ec.c b/src/object/cli_ec.c
@@ -536,7 +536,9 @@ obj_ec_stripe_encode(daos_iod_t *iod, d_sg_list_t *sgl, uint32_t iov_idx,
 		}
 	}
 
-	ec_encode_data(cell_bytes, k, p, codec->ec_gftbls, data, parity_bufs);
+	(void)data;
+	(void)p;
+	//ec_encode_data(cell_bytes, k, p, codec->ec_gftbls, data, parity_bufs);
 
 out:
 	for (i = 0; i < c_idx; i++)

diff --git a/src/object/srv_ec_aggregate.c b/src/object/srv_ec_aggregate.c
@@ -2667,6 +2667,11 @@ cont_ec_aggregate_cb(struct ds_cont_child *cont, daos_epoch_range_t *epr,
 	struct ec_agg_param	 *ec_agg_param = agg_param->ap_data;
 	vos_iter_param_t	 iter_param = { 0 };
 	struct vos_iter_anchors  anchors = { 0 };
+	struct dtx_handle	*dth = NULL;
+	struct dtx_share_peer	*dsp;
+	struct dtx_id		 dti = { 0 };
+	struct dtx_epoch	 epoch = { 0 };
+	daos_unit_oid_t		 oid = { 0 };
 	int			 rc = 0;
 	int                       blocks       = 0;
 
@@ -2715,8 +2720,25 @@ cont_ec_aggregate_cb(struct ds_cont_child *cont, daos_epoch_range_t *epr,
 	agg_reset_entry(&ec_agg_param->ap_agg_entry, NULL, NULL);
 
 retry:
+	epoch.oe_value = epr->epr_hi;
+	rc = dtx_begin(cont->sc_hdl, &dti, &epoch, 0, cont->sc_pool->spc_map_version, &oid,
+		       NULL, 0, 0, NULL, &dth);
+	if (rc != 0)
+		goto update_hae;
+
 	rc = vos_iterate(&iter_param, VOS_ITER_OBJ, true, &anchors, agg_iterate_pre_cb,
-			 agg_iterate_post_cb, ec_agg_param, NULL);
+			 agg_iterate_post_cb, ec_agg_param, dth);
+	if (rc == -DER_INPROGRESS && !d_list_empty(&dth->dth_share_tbd_list)) {
+		while ((dsp = d_list_pop_entry(&dth->dth_share_tbd_list,
+					       struct dtx_share_peer, dsp_link)) != NULL) {
+			D_ERROR(DF_CONT ": EC aggregate hit non-committed DTX " DF_DTI "\n",
+				DP_CONT(cont->sc_pool->spc_uuid, cont->sc_uuid),
+				DP_DTI(&dsp->dsp_xid));
+			dtx_dsp_free(dsp);
+		}
+	}
+
+	dtx_end(dth, cont, rc);
 
 	/* Post_cb may not being executed in some cases */
 	agg_clear_extents(&ec_agg_param->ap_agg_entry);