[
https://issues.apache.org/jira/browse/IGNITE-24060?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Kirill Gusakov updated IGNITE-24060:
------------------------------------
Description:
During the IGNITE-24018 we discovered, that scale down is not working for HA
zones. The root cause is the following code in
[DistributionZoneManager|https://github.com/apache/ignite-3/blob/main/modules/distribution-zones/src/main/java/org/apache/ignite/internal/distributionzones/DistributionZoneManager.java#L980]
{code:java}
if (nodesRemoved) {
if (zone.consistencyMode() == HIGH_AVAILABILITY) {
if (partitionReset != INFINITE_TIMER_VALUE) {
zonesState.get(zoneId).reschedulePartitionDistributionReset(
partitionReset,
() -> fireTopologyReduceLocalEvent(revision, zoneId),
zoneId
);
}
} else {
if (autoAdjustScaleDown == IMMEDIATE_TIMER_VALUE) {
futures.add(saveDataNodesToMetaStorageOnScaleDown(zoneId,
revision));
}
if (autoAdjustScaleDown != INFINITE_TIMER_VALUE) {
zonesState.get(zoneId).rescheduleScaleDown(
autoAdjustScaleDown,
() -> saveDataNodesToMetaStorageOnScaleDown(zoneId,
revision),
zoneId
);
}
}
}
{code}
*Definition of done*
* Scaledown timers are working for HA case also and appropriate tests added
> Scale down timer are not working for HA zones
> ---------------------------------------------
>
> Key: IGNITE-24060
> URL: https://issues.apache.org/jira/browse/IGNITE-24060
> Project: Ignite
> Issue Type: Bug
> Reporter: Kirill Gusakov
> Assignee: Kirill Gusakov
> Priority: Major
> Labels: ignite-3
>
> During the IGNITE-24018 we discovered, that scale down is not working for HA
> zones. The root cause is the following code in
> [DistributionZoneManager|https://github.com/apache/ignite-3/blob/main/modules/distribution-zones/src/main/java/org/apache/ignite/internal/distributionzones/DistributionZoneManager.java#L980]
> {code:java}
> if (nodesRemoved) {
> if (zone.consistencyMode() == HIGH_AVAILABILITY) {
> if (partitionReset != INFINITE_TIMER_VALUE) {
> zonesState.get(zoneId).reschedulePartitionDistributionReset(
> partitionReset,
> () -> fireTopologyReduceLocalEvent(revision, zoneId),
> zoneId
> );
> }
> } else {
> if (autoAdjustScaleDown == IMMEDIATE_TIMER_VALUE) {
> futures.add(saveDataNodesToMetaStorageOnScaleDown(zoneId,
> revision));
> }
> if (autoAdjustScaleDown != INFINITE_TIMER_VALUE) {
> zonesState.get(zoneId).rescheduleScaleDown(
> autoAdjustScaleDown,
> () -> saveDataNodesToMetaStorageOnScaleDown(zoneId,
> revision),
> zoneId
> );
> }
> }
> }
> {code}
> *Definition of done*
> * Scaledown timers are working for HA case also and appropriate tests added
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
