You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Vladislav Pyatkov (Jira)" <ji...@apache.org> on 2023/12/28 11:12:00 UTC
[jira] [Updated] (IGNITE-21061) Durable cleanup requires additional replication group command

     [ https://issues.apache.org/jira/browse/IGNITE-21061?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Vladislav Pyatkov updated IGNITE-21061:
---------------------------------------
    Description: 
h3. Motivation
After locks are released, it is required to write the information to transaction persistent storage and replicate it on all commit partition replication group nodes. That is performed by the replication command ({{MarkLocksReleasedCommand}}). As a result, we have an additional replication command for the entire transaction process.

h3. Implementation notes
In my opinion, we can resolve this situation in the transaction resolution procedure ({{OrphanDetector}}). It just needs to do nothing additional: either release the locks by durable finish from the transaction coordinator or release them by recovery by {{OrphanDetector}}.

  was:
h3. Motivation
After locks are released, it is required to write the information to transaction persistent storage and replicate it on all commit partition replication group nodes. That is performed by the replication command ({{MarkLocksReleasedCommand}}). As a result, we have an additional replication command for the entire transaction process.

h3. Implementation notes
In my opinion, we can resolve this situation in the transaction resolution procedure ({{OrphanDetector}}). It just needs to check the lease on the commit partition. If the lease is changed when we are faced with a transaction lock, the replication process should start.


> Durable cleanup requires additional replication group command
> -------------------------------------------------------------
>
>                 Key: IGNITE-21061
>                 URL: https://issues.apache.org/jira/browse/IGNITE-21061
>             Project: Ignite
>          Issue Type: Improvement
>            Reporter: Vladislav Pyatkov
>            Priority: Major
>              Labels: ignite-3
>
> h3. Motivation
> After locks are released, it is required to write the information to transaction persistent storage and replicate it on all commit partition replication group nodes. That is performed by the replication command ({{MarkLocksReleasedCommand}}). As a result, we have an additional replication command for the entire transaction process.
> h3. Implementation notes
> In my opinion, we can resolve this situation in the transaction resolution procedure ({{OrphanDetector}}). It just needs to do nothing additional: either release the locks by durable finish from the transaction coordinator or release them by recovery by {{OrphanDetector}}.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)