[
https://issues.apache.org/jira/browse/AURORA-111?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Kevin Sweeney updated AURORA-111:
---------------------------------
Description: Right now, whenever a leader election is the newly-elected
leader has to play back the log and rebuild its state, all the while being
offline (the more task state transitions it has to play back, the longer this
takes). Consider keeping this state available on the followers (consistent, but
arbitrarily far behind) and then catching up after a successful leader
election. (was: Right now, whenever a leader election is triggered all
followers dump their state and restart, then the newly-elected leader has to
play back the log and rebuild its state, all the while being offline (the more
task state transitions it has to play back, the longer this takes). Consider
keeping this state available on the followers (consistent, but arbitrarily far
behind) and then catching up after a successful leader election.)
> Support warm failover in schedulers
> -----------------------------------
>
> Key: AURORA-111
> URL: https://issues.apache.org/jira/browse/AURORA-111
> Project: Aurora
> Issue Type: Story
> Components: Scheduler
> Reporter: Kevin Sweeney
> Priority: Minor
>
> Right now, whenever a leader election is the newly-elected leader has to play
> back the log and rebuild its state, all the while being offline (the more
> task state transitions it has to play back, the longer this takes). Consider
> keeping this state available on the followers (consistent, but arbitrarily
> far behind) and then catching up after a successful leader election.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
