[
https://issues.apache.org/jira/browse/IGNITE-16365?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Denis Chudov updated IGNITE-16365:
----------------------------------
Description:
The key point of local state recovery is catching of latest revision applied in
metastorage to apply it locally. In case of standalone mode the node should
finish recovery after restoring components' state from vault, otherwise it
should retrieve updates from metastorage. Notifications from metastorage should
be allowed on recovery stage after all components have started, and recovery
should continue until the node has reached minimal acceptable difference
between itself and the cluster.
Need to implement:
* methods for getting the minimal available revision in Metastorage and the
last applied one;
* some recovery processor class intended to do the logic related to catching
last metastorage revision, and responsile for firing FinishedRecovery message
to cluster management group.
was:
The key point of local state recovery is catching of latest revision applied in
metastorage to apply it locally. In case of standalone mode the node should
finish recovery after restoring components' state from vault, otherwise it
should retrieve updates from metastorage. Notifications from metastorage should
be allowed on recovery stage after all components have started, and recovery
should continue until the node has reached minimal acceptable difference
between itself and the cluster.
Need to implement methods for getting the minimal available revision in
Metastorage and the last applied.
> Implement a logic of recovery finishing
> ---------------------------------------
>
> Key: IGNITE-16365
> URL: https://issues.apache.org/jira/browse/IGNITE-16365
> Project: Ignite
> Issue Type: Improvement
> Reporter: Denis Chudov
> Priority: Major
> Labels: ignite-3
>
> The key point of local state recovery is catching of latest revision applied
> in metastorage to apply it locally. In case of standalone mode the node
> should finish recovery after restoring components' state from vault,
> otherwise it should retrieve updates from metastorage. Notifications from
> metastorage should be allowed on recovery stage after all components have
> started, and recovery should continue until the node has reached minimal
> acceptable difference between itself and the cluster.
> Need to implement:
> * methods for getting the minimal available revision in Metastorage and the
> last applied one;
> * some recovery processor class intended to do the logic related to catching
> last metastorage revision, and responsile for firing FinishedRecovery message
> to cluster management group.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
