Details
-
Bug
-
Resolution: Unresolved
-
Critical
-
3.0.8, 3.0.9, 3.0.10
-
None
-
None
-
1
Description
Running with 3.0.10, 4 kv nodes, 2 replicas, 1 doc loaded, 2 python clients upserting to the same doc concurrently with persist majority durability level. We expect to be able to catch DurabilitySyncWriteInProgressException: https://github.com/couchbase/couchbase-python-client/blob/3.0.10/couchbase/exceptions.py#L1208
However it is only catchable as a CouchbaseException. Furthermore, once we hit competing sync writes and catch the error as a CouchbaseException, LCB will enter an infinite loop of trying to reconnect and retry the op which throws a couchbase.exceptions.LCB_0x401 (LCB_ERR_SOCKET_SHUTDOWN) exception which is caught as a CouchbaseException. After the first sync write in progress error, we will get the LCB_ERR_SOCKET_SHUTDOWN every 2 minutes with no ops succeeding. 2 minutes is the kv timeout set in the test. Memcached rejects the reconnect/retry as "Invalid packet header detected"
Job: http://perf.jenkins.couchbase.com/job/ares/20443/consoleFull
https://s3.amazonaws.com/bugdb/jira/qe/collectinfo-2021-03-26T164016-ns_1%40172.23.133.11.zip
https://s3.amazonaws.com/bugdb/jira/qe/collectinfo-2021-03-26T164016-ns_1%40172.23.133.12.zip
https://s3.amazonaws.com/bugdb/jira/qe/collectinfo-2021-03-26T164016-ns_1%40172.23.133.13.zip
https://s3.amazonaws.com/bugdb/jira/qe/collectinfo-2021-03-26T164016-ns_1%40172.23.133.14.zip