You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Filipp Shergalis (Jira)" <ji...@apache.org> on 2025/12/05 06:47:00 UTC
[jira] [Updated] (IGNITE-27258) Wait for all nodes in MULTI_NODE disaster recovery requests
[ https://issues.apache.org/jira/browse/IGNITE-27258?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Filipp Shergalis updated IGNITE-27258:
--------------------------------------
Description:
Disaster recovery MULTI_NODE requests finish when coordinator node finished processing. We don't wait for other nodes, also we won't even know if other nodes failed to process the request
*What was done:* Every node writes a local request processing state to metastorage (IN_PROGRESS / text of exception / COMPLETED). Coordinator's node has a prefix watch for each operation that is notified of these changes and finishes disaster recovery request processing future only when all nodes report completion or any node fails / leaves the cluster.
Timeout was introduced to fail processing future if any of the nodes didn't start processing request in time. It is hardcoded to 10 seconds
> Wait for all nodes in MULTI_NODE disaster recovery requests
> ------------------------------------------------------------
>
> Key: IGNITE-27258
> URL: https://issues.apache.org/jira/browse/IGNITE-27258
> Project: Ignite
> Issue Type: Improvement
> Reporter: Filipp Shergalis
> Assignee: Filipp Shergalis
> Priority: Major
> Labels: ignite-3
> Time Spent: 20m
> Remaining Estimate: 0h
>
> Disaster recovery MULTI_NODE requests finish when coordinator node finished processing. We don't wait for other nodes, also we won't even know if other nodes failed to process the request
>
> *What was done:* Every node writes a local request processing state to metastorage (IN_PROGRESS / text of exception / COMPLETED). Coordinator's node has a prefix watch for each operation that is notified of these changes and finishes disaster recovery request processing future only when all nodes report completion or any node fails / leaves the cluster.
> Timeout was introduced to fail processing future if any of the nodes didn't start processing request in time. It is hardcoded to 10 seconds
--
This message was sent by Atlassian Jira
(v8.20.10#820010)