Igor Belyakov created IGNITE-16978:
--------------------------------------
Summary: SQL update fails with "modified concurrently" error for
REPLICATED cache with TRANSACTIONAL atomicity mode
Key: IGNITE-16978
URL: https://issues.apache.org/jira/browse/IGNITE-16978
Project: Ignite
Issue Type: Bug
Affects Versions: 2.12
Reporter: Igor Belyakov
In case SQL update query executed for the same key multiple times using JDBC or
Java thin client, sometimes it leads to:
Failed to update some keys because they had been modified concurrently
[keys=[1]]
There is no concurrent updates execution, all updates are executed one by one
in the same thread.
The issue is reproducible only in the case of {{REPLICATED}} cache with
{{TRANSACTIONAL}} atomicity mode and the client should be connected to the
server node where backup entry is stored.
The issue can be easily reproduced if any client node joins/leaves the cluster
during the query execution, but also the issue happens without any additional
actions.
Also, in case {{REPLICATED}} cache is replaced by {{PARTITIONED}} with
{{{}BACKUPS=1{}}}, using a cluster with 2 server nodes, the issue is not
reproducible anymore. The same happens after changing atomicity mode from
{{TRANSACTIONAL}} to {{{}ATOMIC{}}}.
Reproducer attached.
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
