DAOS-17001 rebuild: when self_heal is set to delay_rebuild, do not re…

…build on exclude delay_rebuild mode should delay the rebuild in all scenarios and not have an exception for target exclusion. Also changed an error message to warn on shard update failure. Shard update failure is normal during a failure, and the message was too frequent. Testing: `dmg pool exclude default-pool --rank 0 --target-idx 4` while write/read workflow was running against a cluster Signed-off-by: Chris Davis <[email protected]>
daos-stack · Jan 29, 2025 · f5fb8d7 · f5fb8d7
1 parent 6a1c698
commit f5fb8d7
Show file tree

Hide file tree

Showing 2 changed files with 5 additions and 5 deletions.
diff --git a/src/object/cli_obj.c b/src/object/cli_obj.c
@@ -1,5 +1,6 @@
 /**
  * (C) Copyright 2016-2024 Intel Corporation.
+ * (C) Copyright 2025 Google LLC
  *
  * SPDX-License-Identifier: BSD-2-Clause-Patent
  */
@@ -5816,7 +5817,7 @@ dc_obj_update(tse_task_t *task, struct dtx_epoch *epoch, uint32_t map_ver,
 
 	rc = obj_update_shards_get(obj, args, map_ver, obj_auxi, &shard, &shard_cnt);
 	if (rc != 0) {
-		D_ERROR(DF_OID" get update shards failure %d\n", DP_OID(obj->cob_md.omd_id), rc);
+		D_WARN(DF_OID" get update shards failure %d\n", DP_OID(obj->cob_md.omd_id), rc);
 		D_GOTO(out_task, rc);
 	}
 

diff --git a/src/pool/srv_pool.c b/src/pool/srv_pool.c
@@ -1,6 +1,7 @@
 /*
  * (C) Copyright 2016-2024 Intel Corporation.
  * (C) Copyright 2025 Hewlett Packard Enterprise Development LP
+ * (C) Copyright 2025 Google LLC
  *
  * SPDX-License-Identifier: BSD-2-Clause-Patent
  */
@@ -7343,7 +7344,7 @@ pool_svc_update_map(struct pool_svc *svc, crt_opcode_t opc, bool exclude_rank,
 
 	entry = daos_prop_entry_get(&prop, DAOS_PROP_PO_SELF_HEAL);
 	D_ASSERT(entry != NULL);
-	if (!(entry->dpe_val & (DAOS_SELF_HEAL_AUTO_REBUILD | DAOS_SELF_HEAL_DELAY_REBUILD))) {
+	if (!(entry->dpe_val & DAOS_SELF_HEAL_AUTO_REBUILD)) {
 		D_DEBUG(DB_MD, "self healing is disabled\n");
 		D_GOTO(out, rc);
 	}
@@ -7360,9 +7361,7 @@ pool_svc_update_map(struct pool_svc *svc, crt_opcode_t opc, bool exclude_rank,
 		D_GOTO(out, rc);
 	}
 
-	if ((entry->dpe_val & DAOS_SELF_HEAL_DELAY_REBUILD) && exclude_rank)
-		delay = -1;
-	else if (daos_fail_check(DAOS_REBUILD_DELAY))
+	if (daos_fail_check(DAOS_REBUILD_DELAY))
 		delay = 5;
 
 	D_DEBUG(DB_MD, "map ver %u/%u\n", map_version ? *map_version : -1,