[
https://issues.apache.org/jira/browse/IGNITE-23885?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mirza Aliev updated IGNITE-23885:
---------------------------------
Description:
h3. Motivation
For the purposes of phase 2 of
[IEP-131|https://cwiki.apache.org/confluence/display/IGNITE/IEP-131%3A+Partition+Majority+Unavailability+Handling],
we have introduced (https://issues.apache.org/jira/browse/IGNITE-23870) the
method for receiving a chain of a assignments. This chain will be used in the
restart node logic and based on this chain node will decide, if raft group
should or should not start on the current node.
In case when on restart node receives only one element in this chain, node must
reuse the current logic for the partition to start, based on the info about
stable and pending keys.
h3. Implementation notes
In the improved start table flow on recovery
(https://issues.apache.org/jira/browse/IGNITE-23874), we must add check for the
size of assignments chain, and if this size is equals to 1, the common logic
for the partition start must be reused.
h3. Definition of done
* When on restart node receives only one element in the assignments chain,
node must reuse the current logic for the partition to start, based on the info
about stable and pending keys.
was:
h3. Motivation
For the purposes of phase 2 of
[IEP-131|https://cwiki.apache.org/confluence/display/IGNITE/IEP-131%3A+Partition+Majority+Unavailability+Handling],
we have introduced (https://issues.apache.org/jira/browse/IGNITE-23870) the
method for receiving a chain of a assignments. This chain will be used in the
restart node logic and based on this chain node will decide, if raft group
should or should not start on the current node.
In case when on restart node receives only one element in this chain, node must
reuse the current logic for the partition to start, based on the info about
stable and pending keys.
h3. Implementation notes
In the improved start table flow
(https://issues.apache.org/jira/browse/IGNITE-23874), we must add check for the
size of assignments chain, and if this size is equals to 1, the common logic
for the partition start must be reused.
h3. Definition of done
* When on restart node receives only one element in the assignments chain,
node must reuse the current logic for the partition to start, based on the info
about stable and pending keys.
> Start partition in a common manner (stable + pending) in case of a single
> element in chain
> ------------------------------------------------------------------------------------------
>
> Key: IGNITE-23885
> URL: https://issues.apache.org/jira/browse/IGNITE-23885
> Project: Ignite
> Issue Type: New Feature
> Reporter: Alexander Lapin
> Priority: Major
>
> h3. Motivation
> For the purposes of phase 2 of
> [IEP-131|https://cwiki.apache.org/confluence/display/IGNITE/IEP-131%3A+Partition+Majority+Unavailability+Handling],
>
> we have introduced (https://issues.apache.org/jira/browse/IGNITE-23870) the
> method for receiving a chain of a assignments. This chain will be used in the
> restart node logic and based on this chain node will decide, if raft group
> should or should not start on the current node.
> In case when on restart node receives only one element in this chain, node
> must reuse the current logic for the partition to start, based on the info
> about stable and pending keys.
> h3. Implementation notes
> In the improved start table flow on recovery
> (https://issues.apache.org/jira/browse/IGNITE-23874), we must add check for
> the size of assignments chain, and if this size is equals to 1, the common
> logic for the partition start must be reused.
> h3. Definition of done
> * When on restart node receives only one element in the assignments chain,
> node must reuse the current logic for the partition to start, based on the
> info about stable and pending keys.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
