Maksim Timonin created IGNITE-18782:
---------------------------------------
Summary: Snapshot restore might fail after cluster become active
due to transition state
Key: IGNITE-18782
URL: https://issues.apache.org/jira/browse/IGNITE-18782
Project: Ignite
Issue Type: Bug
Reporter: Maksim Timonin
`ign.cluster().state(ClusterState.ACTIVE)` activates cluster but after this
method returns some grid nodes might be in transition state -
`ctx.state().clusterState().transition() == true`.
Ignite nodes checks the transition state before restoring snapshot, and then
restoring procedure might fail.
Also, it looks like `GridClusterStateProcessor#publicApiState(boolean
waitForTransition)` doesn't check transition state. Example is
IgniteClusterSnapshotRestoreSelfTest#testUserClassRestored - it's required to
explicitly check transition to start restoring snapshot, usage of the
publicApiState call doesn't help. Maybe bug here?
There are some possible ways to fix it:
# `state(ClusterState.ACTIVE)` should return after transition has fully
completed
# Snapshot restoring should await while transition state finished.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
