[ 
https://issues.apache.org/jira/browse/IGNITE-23879?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Mirza Aliev updated IGNITE-23879:
---------------------------------
    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],
  
and https://issues.apache.org/jira/browse/IGNITE-23870, 

we need to have an ability to understand in the stable switch phase, if 
planned, that is going to be written to pending, is triggered by two phase 
resetPartitions, or it is is triggered by graceful change, like scale up and 
etc. This is needed because we want to decide in the 
https://issues.apache.org/jira/browse/IGNITE-23870, should we rewrite newly 
introduced key {{assignments.chain}} with current pending or need to add this 
pending to the chain. Details could be found here 
(https://issues.apache.org/jira/browse/IGNITE-23870)

h3. Implementation notes
We introduce new boolean filed in {{partitiondistribution.Assignments}}, let's 
name it like {{secondPhaseOfReset}}

In {{GroupUpdateRequest#partitionUpdate}}, when we write planned for the second 
phase of reset, we must set {{planned.secondPhaseOfReset(true)}}
 

h3. Definition of done
* Planned assignment in the second phase of reset is enriched with  
{{secondPhaseOfReset(true)}}
* Pending after stable switch, which was written from that pending, also must 
have {{secondPhaseOfReset == true}}
* Asserts must be checked:
{noformat}
            // force == true, asserts secondPhaseOfReset == false 
            // secondPhaseOfReset == true, assert force == false 
{noformat}



  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],
  
and https://issues.apache.org/jira/browse/IGNITE-23870, 

we need to have an ability to understand in the stable switch phase, if 
planned, that is going to be written to pending, is triggered by two phase 
resetPartitions, or it is is triggered by graceful change, like scale up and 
etc. This is needed because we want to decide in the 
https://issues.apache.org/jira/browse/IGNITE-23870, should we rewrite newly 
introduced key {{assignments.chain}} with current pending or need to add this 
pending to the chain. Details could be found here 
(https://issues.apache.org/jira/browse/IGNITE-23870)

h3. Implementation notes
We introduce new boolean filed in {{partitiondistribution.Assignments}}, let's 
name it like {{secondPhaseOfReset}}

In {{GroupUpdateRequest#partitionUpdate}}, when we write planned for the second 
phase of reset, we must set {{planned.secondPhaseOfReset(true)}}
 

h3. Definition of done
* Planned assignment in the second phase of reset is enriched with  
{{secondPhaseOfReset(true)}}
* Pending after stable switch, which was written from that pending, also must 
have {{secondPhaseOfReset == true}}
* Asserts must be checked:
{noformat}
            // force = true, asserts secondPhaseOfReset == false 
            // secondPhaseOfReset == true, assert force == false 
{noformat}




> Provide a way to indicate stable switch after the second phase of 
> resetPartitions
> ---------------------------------------------------------------------------------
>
>                 Key: IGNITE-23879
>                 URL: https://issues.apache.org/jira/browse/IGNITE-23879
>             Project: Ignite
>          Issue Type: Improvement
>            Reporter: Mirza Aliev
>            Priority: Major
>              Labels: ignite-3
>
> 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],
>   
> and https://issues.apache.org/jira/browse/IGNITE-23870, 
> we need to have an ability to understand in the stable switch phase, if 
> planned, that is going to be written to pending, is triggered by two phase 
> resetPartitions, or it is is triggered by graceful change, like scale up and 
> etc. This is needed because we want to decide in the 
> https://issues.apache.org/jira/browse/IGNITE-23870, should we rewrite newly 
> introduced key {{assignments.chain}} with current pending or need to add this 
> pending to the chain. Details could be found here 
> (https://issues.apache.org/jira/browse/IGNITE-23870)
> h3. Implementation notes
> We introduce new boolean filed in {{partitiondistribution.Assignments}}, 
> let's name it like {{secondPhaseOfReset}}
> In {{GroupUpdateRequest#partitionUpdate}}, when we write planned for the 
> second phase of reset, we must set {{planned.secondPhaseOfReset(true)}}
>  
> h3. Definition of done
> * Planned assignment in the second phase of reset is enriched with  
> {{secondPhaseOfReset(true)}}
> * Pending after stable switch, which was written from that pending, also must 
> have {{secondPhaseOfReset == true}}
> * Asserts must be checked:
> {noformat}
>             // force == true, asserts secondPhaseOfReset == false 
>             // secondPhaseOfReset == true, assert force == false 
> {noformat}



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

Reply via email to