[
https://issues.apache.org/jira/browse/IGNITE-21370?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Roman Puchkovskiy updated IGNITE-21370:
---------------------------------------
Summary: Handle unacknowledged messages queue overflow (was: Handle
unacknowledge messages queue overflow)
> Handle unacknowledged messages queue overflow
> ---------------------------------------------
>
> Key: IGNITE-21370
> URL: https://issues.apache.org/jira/browse/IGNITE-21370
> Project: Ignite
> Issue Type: Improvement
> Reporter: Roman Puchkovskiy
> Priority: Major
> Labels: ignite-3
> Fix For: 3.0.0-beta2
>
>
> An unacknowledges messages queue contained in RecoveryDescriptor is bounded.
> Right now, its overflow (that causes an exception to be thrown) is not
> handled.
> When the queue becomes full, we could forcibly close the cnannel to force
> reconnection and unacked messages being resent during a handshake. We need to
> be careful with message ordering wrt the message that causes the message
> queue to be full. It seems that we need to detect 'the queue has JUST become
> full, after adding this message', and not 'we cannot add this message because
> the queue is full' as in the latter case it would be difficult to guarantee
> the ordering (as the message is not added to the queue).
> Also, currently max queue size is 10. A better default should be chosen, and
> this must be made configurable.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
