DAOS-14010 rebuild: discard non-exist cont/obj for reint

Some containers/objects possibly be destroyed between exclude and reint, so during reintegration should try to discard those containers or objects. Add reint post process for that purpose. Set DAOS_INCR_REINT_ENABLE ENV as "1" to enable incremental reint, by default it is disabled for 2.8. Skip-nlt: true Signed-off-by: Xuezhao Liu <[email protected]>
daos-stack · Jan 12, 2025 · a92a767 · a92a767
1 parent 14e3039
commit a92a767
Show file tree

Hide file tree

Showing 10 changed files with 398 additions and 93 deletions.
diff --git a/src/container/srv_container.c b/src/container/srv_container.c
@@ -1832,7 +1832,7 @@ int
 ds_cont_tgt_refresh_track_eph(uuid_t pool_uuid, uuid_t cont_uuid,
 			      daos_epoch_t ec_agg_eph, daos_epoch_t stable_eph)
 {
-	struct ds_pool			*pool;
+	struct ds_pool			*pool = NULL;
 	struct pool_target		*tgts;
 	struct refresh_track_eph_arg	 arg;
 	d_rank_t			 rank;
@@ -1861,17 +1861,16 @@ ds_cont_tgt_refresh_track_eph(uuid_t pool_uuid, uuid_t cont_uuid,
 	rank = dss_self_rank();
 	rc = pool_map_find_target_by_rank_idx(pool->sp_map, rank, -1, &tgts);
 	D_ASSERT(rc == dss_tgt_nr);
-	for (i = 0; i < dss_tgt_nr; i++) {
+	for (i = 0; i < dss_tgt_nr; i++)
 		arg.tgt_status[i] = tgts[i].ta_comp.co_status;
-	}
 	ds_pool_put(pool);
 
 	rc = ds_pool_task_collective(pool_uuid, PO_COMP_ST_NEW | PO_COMP_ST_DOWN |
 				     PO_COMP_ST_DOWNOUT, cont_refresh_track_eph_one,
 				     &arg, DSS_ULT_FL_PERIODIC);
 
 out:
-	if (arg.tgt_status != arg.tgt_status_inline)
+	if (arg.tgt_status != NULL && arg.tgt_status != arg.tgt_status_inline)
 		D_FREE(arg.tgt_status);
 	return rc;
 }

diff --git a/src/include/daos_srv/daos_engine.h b/src/include/daos_srv/daos_engine.h
@@ -1,5 +1,5 @@
 /**
- * (C) Copyright 2016-2024 Intel Corporation.
+ * (C) Copyright 2016-2025 Intel Corporation.
  *
  * SPDX-License-Identifier: BSD-2-Clause-Patent
  */
@@ -694,9 +694,9 @@ unsigned int dss_ctx_nr_get(void);
 
 /* Cache for container root */
 struct tree_cache_root {
-	struct btr_root	btr_root;
-	daos_handle_t	root_hdl;
-	unsigned int	count;
+	struct btr_root	tcr_btr_root;
+	daos_handle_t	tcr_root_hdl;
+	unsigned int	tcr_count;
 };
 
 int

diff --git a/src/include/daos_srv/object.h b/src/include/daos_srv/object.h
@@ -1,5 +1,6 @@
 /*
- * (C) Copyright 2022 Intel Corporation.
+ * (C) Copyright 2022-2025 Intel Corporation.
+ * (C) Copyright 2025 Hewlett Packard Enterprise Development LP
  *
  * SPDX-License-Identifier: BSD-2-Clause-Patent
  */
@@ -76,7 +77,7 @@ struct ds_migrate_status {
 
 int
 ds_migrate_query_status(uuid_t pool_uuid, uint32_t ver, unsigned int generation, int op,
-			struct ds_migrate_status *dms);
+			bool gl_scan_done, struct ds_migrate_status *dms);
 
 int
 ds_object_migrate_send(struct ds_pool *pool, uuid_t pool_hdl_uuid, uuid_t cont_uuid,

diff --git a/src/include/daos_srv/pool.h b/src/include/daos_srv/pool.h
@@ -395,6 +395,7 @@ void ds_pool_disable_exclude(void);
 void ds_pool_enable_exclude(void);
 
 extern bool ec_agg_disabled;
+extern bool daos_incr_reint_enabled;
 
 int dsc_pool_open(uuid_t pool_uuid, uuid_t pool_hdl_uuid,
 		       unsigned int flags, const char *grp,

diff --git a/src/object/srv_internal.h b/src/object/srv_internal.h
@@ -102,8 +102,8 @@ struct migrate_pool_tls {
 	unsigned int		mpt_ult_running:1,
 				mpt_init_tls:1,
 				mpt_fini:1,
-				mpt_reintegrating_check:1,
-				mpt_reintegrating:1;
+				mpt_reintegrating:1, /* incremental reint flag */
+				mpt_post_process_started:1; /* reint post process started flag */
 
 	/* migration init error */
 	int			mpt_init_err;