[jira] [Commented] (AURORA-1812) Upgrading scheduler multiple times in succession can lead to incompatible snapshot restore

["Zameer Manji (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/AURORA-1812?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15652255#comment-15652255 ] 

Zameer Manji commented on AURORA-1812:
--------------------------------------

[~joshua.cohen] [~StephanErb]

Maybe we can fix this by having the scheduler take a (new) snapshot right after recovery if there was schema migrations?

> Upgrading scheduler multiple times in succession can lead to incompatible snapshot restore 
> -------------------------------------------------------------------------------------------
>
>                 Key: AURORA-1812
>                 URL: https://issues.apache.org/jira/browse/AURORA-1812
>             Project: Aurora
>          Issue Type: Bug
>          Components: Scheduler
>    Affects Versions: 0.14.0
>         Environment: Mesos-0.27.2 aurora-scheduler-0.14.0
>            Reporter: Patrick Veasey
>            Priority: Minor
>
> When upgrading scheduler multiple times in a row there can be a situation where the snapshot is restored is from an incompatible version. Which will cause scheduler to fail to start, with SQL exceptions. Workaround is to ensure the most current snapshot was taken by the current version of aurora, either by manually starting snapshot or setting dlog_snapshot_interval to a low timeframe. 
> Log of failure can be found here:
> https://gist.github.com/Pveasey/4ca1ad4d3ded21cd6e1674f20a8a4af3



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)