[
https://issues.apache.org/jira/browse/IGNITE-24056?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Roman Puchkovskiy updated IGNITE-24056:
---------------------------------------
Description:
If a Raft state machine (represented with a RaftGroupListener) starts executing
a batch of commands which is huge, there is no way to stop it midway (apart
from closing a storage it interacts with). This, among other things, does not
allow us to stop an Ignite node timely if some Raft node recovers from a huge
Raft log.
Here is how we can do it:
# As soon as Raft node's shutdown is initiated, make the currently executed
iterator return false from hasNext()
# When such an iterator 'finishes', complete all command closures that were
not handled due to an early stop with a ShutdownException
> Prompt stop of Raft state machines
> ----------------------------------
>
> Key: IGNITE-24056
> URL: https://issues.apache.org/jira/browse/IGNITE-24056
> Project: Ignite
> Issue Type: Improvement
> Reporter: Roman Puchkovskiy
> Assignee: Roman Puchkovskiy
> Priority: Major
> Labels: ignite-3
>
> If a Raft state machine (represented with a RaftGroupListener) starts
> executing a batch of commands which is huge, there is no way to stop it
> midway (apart from closing a storage it interacts with). This, among other
> things, does not allow us to stop an Ignite node timely if some Raft node
> recovers from a huge Raft log.
> Here is how we can do it:
> # As soon as Raft node's shutdown is initiated, make the currently executed
> iterator return false from hasNext()
> # When such an iterator 'finishes', complete all command closures that were
> not handled due to an early stop with a ShutdownException
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
