[
https://issues.apache.org/jira/browse/IGNITE-24892?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mirza Aliev updated IGNITE-24892:
---------------------------------
Description:
h3. Motivation
The Strongly Consistent (SC) mode of a zone ensures data consistency using the
Raft algorithm. However, if a majority of nodes hosting a partition are lost,
the SC partition becomes unavailable.
Due to the guarantees provided by the Raft algorithm, we might assume that in a
two-node scenario, if data has been applied to Raft, it has also been applied
to both nodes. In this case, even if one of the two nodes is lost, restoring
the partition’s availability would not result in data loss. This suggests that
in a two-node setup, it is possible to recover partition availability in SC
mode, even when the majority is lost.
h3. Definition of Done
* The solution is implemented so that, in a two-node scenario, partition
availability can be recovered without data loss.
was:
h3. Motivation
Strongly Consistent (SC) mode of a zone guaranties data consistency using Raft
algorithm. Loosing majority of nodes hosting partition means that SC partition
becomes unavailable.
Because of the Raft algorithm guaranties, we could believe that in two nodes
scenario if data has been applied to Raft, that means that data is applied to
both nodes.
In that case in SC mode, even if lose one node, if we could recover
availability of a partition, we won't lose data.
> Support availability of SC zone when one node left in two-nodes scenario
> ------------------------------------------------------------------------
>
> Key: IGNITE-24892
> URL: https://issues.apache.org/jira/browse/IGNITE-24892
> Project: Ignite
> Issue Type: Improvement
> Reporter: Mirza Aliev
> Priority: Major
> Labels: ignite-3
>
> h3. Motivation
> The Strongly Consistent (SC) mode of a zone ensures data consistency using
> the Raft algorithm. However, if a majority of nodes hosting a partition are
> lost, the SC partition becomes unavailable.
> Due to the guarantees provided by the Raft algorithm, we might assume that in
> a two-node scenario, if data has been applied to Raft, it has also been
> applied to both nodes. In this case, even if one of the two nodes is lost,
> restoring the partition’s availability would not result in data loss. This
> suggests that in a two-node setup, it is possible to recover partition
> availability in SC mode, even when the majority is lost.
> h3. Definition of Done
> * The solution is implemented so that, in a two-node scenario, partition
> availability can be recovered without data loss.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
