Aleksandr Chesnokov created IGNITE-29135:
--------------------------------------------
Summary: If partitionLossPolicy in nodes' caches configurations
differ, a node configured with READ_WRITE_SAFE reads lost partitions without an
exception
Key: IGNITE-29135
URL: https://issues.apache.org/jira/browse/IGNITE-29135
Project: Ignite
Issue Type: Bug
Reporter: Aleksandr Chesnokov
If a statically configured cache has different partitionLossPolicy on different
nodes, the joining node is accepted without any warning and the coordinator's
policy is applied to the whole cluster, although getConfiguration() on each
node still returns the node's own value. So a node started with READ_WRITE_SAFE
silently reads from lost partitions when the coordinator was started with
IGNORE. For caches inside a cache group such a mismatch is already rejected by
validateCacheGroupsAttributesMismatch, but for standalone caches the attribute
is not checked at all
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
