[
https://issues.apache.org/jira/browse/SOLR-8933?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15239738#comment-15239738
]
Mark Miller commented on SOLR-8933:
-----------------------------------
bq. ServletInputStreamWrapper
Let's add a class comment that mentions why this exists.
bq. + if (System.getProperty("jetty.testMode") != null && !retry) {
I really don't like introducing any memory barriers for this at such an
important code path. I wonder if we could use an assert and private method
somehow.
It seems that even if it works, it's less than ideal to make it a static member
and count on initialize vs setting the test sys prop order, but perhaps that is
an option as well?
bq. new CloseShieldInputStream
I think it's worth adding a small comment when we use this that says something
like: protect container owned streams from being closed by Solr, see SOLR-8933
bq. log.warn("", new IOException("Attempted close of request input stream."));
Since this is test logging, perhaps it should be ERROR so that violators feel
worse about it if they see the logging. Warnings are easy to ignore.
> SolrDispatchFilter::consumeInput logs "Stream Closed" IOException
> -----------------------------------------------------------------
>
> Key: SOLR-8933
> URL: https://issues.apache.org/jira/browse/SOLR-8933
> Project: Solr
> Issue Type: Bug
> Affects Versions: 4.10.3
> Reporter: Mike Drob
> Assignee: Mark Miller
> Attachments: SOLR-8933.patch, SOLR-8933.patch, SOLR-8933.patch,
> SOLR-8933.patch, SOLR-8933.patch, SOLR-8933.patch, SOLR-8933.patch
>
>
> After SOLR-8453 we started seeing some IOExceptions coming out of
> SolrDispatchFilter with "Stream Closed" messages.
> It looks like we are indeed closing the request stream in several places when
> we really need to be letting the web container handle their life cycle. I've
> got a preliminary patch ready and am working on testing it to make sure there
> are no regressions.
> A very strange piece of this is that I have been entirely unable to reproduce
> it on a unit test, but have seen it on cluster deployment quite consistently.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
