[
https://issues.apache.org/jira/browse/BEAM-31?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15263083#comment-15263083
]
Mark Shields commented on BEAM-31:
----------------------------------
A customer hit this issue and it lead to catastrophic data loss: the old
pipeline was killed by --update, the new pipeline came up but with inconsistent
interpretation of trigger state, eventually the new pipeline threw. Even
reverting the trigger code and --updating would not work because the pipeline
was stuck in work item retries and had partially proceeded with corrupted
trigger state.
> When triggers are changed via pipeline update, stale finished triggers data
> applied
> -----------------------------------------------------------------------------------
>
> Key: BEAM-31
> URL: https://issues.apache.org/jira/browse/BEAM-31
> Project: Beam
> Issue Type: Bug
> Components: runner-core
> Reporter: Kenneth Knowles
> Labels: Triggers, Update
>
> TriggerRunner tracks which trigger subexpresions are finished. When the
> trigger expession is updated via pipeline update, this data is applied
> arbitrarily to the new trigger expression.
> Implementation note: trigger subexpressions are identified by number, and
> their finished state stored in a bit set.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
