[jira] [Updated] (IGNITE-19467) Remove redundant zones' iteration on DistributionManager.start and optimize usage of cfg.change in the DistributionManager

["Mirza Aliev (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/IGNITE-19467?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Mirza Aliev updated IGNITE-19467:
---------------------------------
    Description: 
*Motivation:*
It DistributionManager.start we have call of initDataNodesFromVaultManager, which contains the following code 

{noformat}
                zonesConfiguration.distributionZones().value().forEach(zone -> {
                    int zoneId = zone.zoneId();

                    saveDataNodesAndUpdateTriggerKeysInMetaStorage(
                            zoneId,
                            appliedRevision,
                            logicalTopology
                    );
                });
{noformat}

This logic is redundant, because we have the same logic on a zone creation ({{DistributionZoneManager.ZonesConfigurationListener#onCreate}}), and when node is restarted, all configuration changes are applied again, meaning that {{DistributionZoneManager.ZonesConfigurationListener#onCreate}} will be called. Also we can review the possibility to remove the same code for a default zone, but instead of the named list scenario with the user's zones, {{onCreate}} for the default zone is not triggered after restart. 

Also there are several places in the code, which we use such constructions:


{noformat}
change = zonesConfiguration.change(zonesChange -> zonesChange.changeDistributionZones(zonesListChange -> {
{noformat}

This code could be optimised and redundant {{zonesChange.changeDistributionZones(zonesListChange ->  }} closure can be eliminated


  was:
*Motivation:*
It DistributionManager.start we have call of initDataNodesFromVaultManager, which contains the following code 

{noformat}
                zonesConfiguration.distributionZones().value().forEach(zone -> {
                    int zoneId = zone.zoneId();

                    saveDataNodesAndUpdateTriggerKeysInMetaStorage(
                            zoneId,
                            appliedRevision,
                            logicalTopology
                    );
                });
{noformat}

This logic is redundant, because we have the same logic on a zone creation ({{DistributionZoneManager.ZonesConfigurationListener#onCreate}}), and when node is restarted, all configuration changes are applied again, meaning that {{DistributionZoneManager.ZonesConfigurationListener#onCreate}} will be called.

Also there are several places in the code, which we use such constructions:


{noformat}
change = zonesConfiguration.change(zonesChange -> zonesChange.changeDistributionZones(zonesListChange -> {
{noformat}

This code could be optimised and redundant {{zonesChange.changeDistributionZones(zonesListChange ->  }} closure can be eliminated



> Remove redundant zones' iteration on DistributionManager.start and optimize usage of cfg.change in the DistributionManager  
> ----------------------------------------------------------------------------------------------------------------------------
>
>                 Key: IGNITE-19467
>                 URL: https://issues.apache.org/jira/browse/IGNITE-19467
>             Project: Ignite
>          Issue Type: Bug
>            Reporter: Mirza Aliev
>            Priority: Major
>              Labels: ignite-3
>
> *Motivation:*
> It DistributionManager.start we have call of initDataNodesFromVaultManager, which contains the following code 
> {noformat}
>                 zonesConfiguration.distributionZones().value().forEach(zone -> {
>                     int zoneId = zone.zoneId();
>                     saveDataNodesAndUpdateTriggerKeysInMetaStorage(
>                             zoneId,
>                             appliedRevision,
>                             logicalTopology
>                     );
>                 });
> {noformat}
> This logic is redundant, because we have the same logic on a zone creation ({{DistributionZoneManager.ZonesConfigurationListener#onCreate}}), and when node is restarted, all configuration changes are applied again, meaning that {{DistributionZoneManager.ZonesConfigurationListener#onCreate}} will be called. Also we can review the possibility to remove the same code for a default zone, but instead of the named list scenario with the user's zones, {{onCreate}} for the default zone is not triggered after restart. 
> Also there are several places in the code, which we use such constructions:
> {noformat}
> change = zonesConfiguration.change(zonesChange -> zonesChange.changeDistributionZones(zonesListChange -> {
> {noformat}
> This code could be optimised and redundant {{zonesChange.changeDistributionZones(zonesListChange ->  }} closure can be eliminated



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