Details
-
Bug
-
Resolution: Fixed
-
Major
-
6.5.0
-
6.5.0-4959-enterprise
-
Untriaged
-
Centos 64-bit
-
No
-
KV Spint 2020-March
Description
Build: 6.5.0-4959
Scenario:
- 2 node cluster, couchbase bucket(replica=1)
- Stop memcached on replica node
- INSERT a new doc into the bucket with timeout=60seconds using client-1 with durability=MAJORITY
- Try REPLACE of same doc from client-2
Observation:
Client-2 gets 'RequestCancelledException' with reason 'KV_SYNC_WRITE_IN_PROGRESS' for the replace request.
[subdoc_xattr:test_doc_sync_write_in_progress:1661] Testing replace over create
|
[sdk_client3:replace:507] Request cancelled/timed-out: com.couchbase.client.core.error.RequestCanceledException: ReplaceRequest {"cancelled":true,"completed":true,"coreId":1,"idempotent":false,"lastChannelId":"0000000000000001/000000000DEDC0B3","lastDispatchedFrom":"10.112.191.1:51579","lastDispatchedTo":"10.112.191.102:11210","reason":"NO_MORE_RETRIES (KV_SYNC_WRITE_IN_PROGRESS)","requestId":51,"requestType":"ReplaceRequest","retried":0,"service":{"bucket":"default","collection":"_default","documentId":"xattrs-0","opaque":"0x5e","scope":"_default","syncDurability":{"present":true},"type":"kv"},"timeoutMs":3000,"timings":{"dispatchMicros":842,"totalMicros":1989}}
|
Expected behavior:
Client-2 should get 'DocumentNotFoundException' since the replace is supposed to get the committed document from the bucket.
Since client-1 sync_write is still in progress, the document should not be exposed to client-2's replace request.
Note:
Currently DELETE over uncommitted document will get 'DocumentNotFoundException'