Map stuck in not ready state, if the master node doesn't have anything to load #1009
Comments
|
That should be fixed in 3.1. LoadAll mechanism has been changed substantially. Can you test with 3.1? If problem exists you can re-open the issue. |
devOpsHazelcast
pushed a commit
that referenced
this issue
Mar 8, 2024
Avoid race conditions between async (via events) proxy initialization on other members than the one who got `InitializeDistributedObjectOperation` and cluster restart. Those races are unfortunate but are not tested here. Fixes HZ-4534, HZ-4536 Fixes: https://github.com/hazelcast/hazelcast-enterprise/issues/7063 Fixes: #26269 Fixes part of: https://github.com/hazelcast/hazelcast-enterprise/issues/6550#issuecomment-1966769964 --------- Co-authored-by: ihsan demir <ihsan@hazelcast.com> GitOrigin-RevId: c33a2000c268af7ff2d03f67e3e266ad637af3e6
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
In hazelcast 3.0.3, there's a bug during MapLoading. If the master (oldest) node doesn't own any keys from the Set loadAllKeys() gave back, the map will be stuck in not ready state forever.
If it is accessed, after a while it will start throwing "RetryableHazelcastException: Map is not ready" exceptions.
This is not a problem when loadAllKeys() give back 0 or many keys, but can happen easily, if it gives back only a few.
The easiest way to recreate is to start many instances, and have loadAllKeys() give back only 1 key.
The text was updated successfully, but these errors were encountered: