Details
-
Bug
-
Status: Closed
-
Critical
-
Resolution: Fixed
-
6.5.0
-
6.5.0-4874
-
Triaged
-
-
Yes
Description
Steps:
1. Initialize a cluster with 2 nodes
2. Create a bucket with replica=1 and load 250 docs
3. Start swap rebalance with 1 node coming in and 1 going out
4. Start doc loading in parallel to step 3
5. Stop rebalance at 20% and start again
6. Stop rebalance at 40% and start again
7. Stop rebalance at 60% and start again, Rebalance failed.
NOTE: Test passed on 6.5.0-4821 and it is reproducible easily for the test given below.
QE Note:
get-cbcollect-info=False,num_items=250000,GROUP=P0;durability,EXCLUDE_GROUP=not_for_majority,durability=MAJORITY,infra_log_level=debug -t rebalance_new.swaprebalancetests.SwapRebalanceStartStopTests.do_test,replicas=1,nodes_init=2,standard_buckets=1,num-swap=1,new_replica=2,GROUP=P0;durability,skip_cleanup=True
|
Attachments
Issue Links
- mentioned in
-
Page Loading...
The changelog (of possible relevant changes) between 6.5.0-4821 and 6.5.0-4874 for KV is as follows:
CHANGELOG for kv_engine
* Commit: 117d7a94723b62d71f3be94e4d72cbb73f0a5797 in build: 6.5.0-4860
MB-36973: Don't use ThreadLocalPtr for CouchKVStore::pendingFileDeletions
* Commit: e1da4853859b19125ac7babba6643e0b44fe0484 in build: 6.5.0-4857
MB-36765: Fix vbucket_state::operator==
* Commit: db516869ffc3d4fa7405f0f11a00542d273b62be in build: 6.5.0-4850
MB-36133: Persist highPreparedSeqno
* Commit: ead473d8324e96eb909938b51b61cfa7e243a919 in build: 6.5.0-4848
MB-36940: Handle tombstones properly in couchfile_upgrade
* Commit: b83516a2c6d9513cac30093d070cb9d530cde287 in build: 6.5.0-4827
MB-36923: Add support for num_reader/writer_threads
* Commit: ad3a4a3f5a93af6bba5829a669bdcd3430bd8147 in build: 6.5.0-4826
MB-36915: Avoid lock-inversion at set-vbstate and new-producer