[
https://issues.apache.org/jira/browse/IGNITE-18681?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Roman Puchkovskiy updated IGNITE-18681:
---------------------------------------
Description:
When a node disappears from the point of view of a discovery mechanism
(currently, SWIM), we employ a delay before removing it from the logical
topology to make sure that short network glitches do not cause topology
instability. But when a CMG leader changes, it first tries to find nodes that
disappeared from the cluster on the SWIM level and removes them from the
logical topology as well. This process ignored the delay.
It should use the same delay probably.
*Update*: this is irrelevant as the delay is removed
was:
When a node disappears from the point of view of a discovery mechanism
(currently, SWIM), we employ a delay before removing it from the logical
topology to make sure that short network glitches do not cause topology
instability. But when a CMG leader changes, it first tries to find nodes that
disappeared from the cluster on the SWIM level and removes them from the
logical topology as well. This process ignored the delay.
It should use the same delay probably.
> Respect removal delay when removing nodes from logical topology after leader
> change
> -----------------------------------------------------------------------------------
>
> Key: IGNITE-18681
> URL: https://issues.apache.org/jira/browse/IGNITE-18681
> Project: Ignite
> Issue Type: Improvement
> Reporter: Roman Puchkovskiy
> Assignee: Roman Puchkovskiy
> Priority: Major
> Labels: ignite-3
> Fix For: 3.2
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> When a node disappears from the point of view of a discovery mechanism
> (currently, SWIM), we employ a delay before removing it from the logical
> topology to make sure that short network glitches do not cause topology
> instability. But when a CMG leader changes, it first tries to find nodes that
> disappeared from the cluster on the SWIM level and removes them from the
> logical topology as well. This process ignored the delay.
> It should use the same delay probably.
> *Update*: this is irrelevant as the delay is removed
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
