Kirill Tkalenko created IGNITE-25502:
----------------------------------------
Summary: Incorrect partition state when entering node with index
greater than current majority
Key: IGNITE-25502
URL: https://issues.apache.org/jira/browse/IGNITE-25502
Project: Ignite
Issue Type: Bug
Reporter: Kirill Tkalenko
When analyzing the test ticket from IGNITE-24802, it was found that *without*
taking a snapshot, when returning a node with a large index, its raft log
suffix is truncated, the partition is in a healthy state and its storages are
not cleared. Thus, we can get records from this node that are not in the
cluster and have no way of knowing about it. This needs to be fixed, but we
need to understand how best to do this.
Perhaps the right decision would be not to truncate the raft log suffix and,
when a conflict between a local index and the majority is detected, transfer
the partition to an erroneous state and give the user the opportunity to fix it
with disaster recovery.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
