[
https://issues.apache.org/jira/browse/LUCENE-6217?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14305029#comment-14305029
]
Robert Muir commented on LUCENE-6217:
-------------------------------------
The way i see it, 100% of users wont catch this. The reason is you cannot
recover from it. If you cannot flush segments or some other error like this,
usually manual intervention is necessary. Its not something you can recover
from.
So, I dont think we need to make big changes to indexwriter or overengineer an
API for it. Changing all exception types to be a fake one definitely counts as
that to me.
> IndexWriter should make it clear when tragedy strikes
> -----------------------------------------------------
>
> Key: LUCENE-6217
> URL: https://issues.apache.org/jira/browse/LUCENE-6217
> Project: Lucene - Core
> Issue Type: Bug
> Reporter: Michael McCandless
> Assignee: Michael McCandless
>
> If you hit an exception at a "bad time" e.g. when writing files for a newly
> flushed segment, IndexWriter declares it a tragedy and secretly closes itself
> as a side effect of the exception.
> Subsequent operations will throw an ACE with the exception that caused the
> tragedy as its cause.
> This requires messy code, if you want to know when this happened to you,
> since the first exception doesn't make it clear that it was "tragic".
> I think we should make it easier to know when this happens?
> Maybe we could instead throw a new exception (IWClosedByTragedy or
> something), or maybe we add a getter (.getTragicException) to IW?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
