You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@geode.apache.org by "ASF subversion and git services (Jira)" <ji...@apache.org> on 2020/09/10 07:12:00 UTC

[jira] [Commented] (GEODE-8053) Hang in legacy integration tests at shutdown

    [ https://issues.apache.org/jira/browse/GEODE-8053?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17193425#comment-17193425 ] 

ASF subversion and git services commented on GEODE-8053:
--------------------------------------------------------

Commit c46f42c732dc095c881d6eddf572ced1e0ecc99a in geode-native's branch refs/heads/master from Jacob Barrett
[ https://gitbox.apache.org/repos/asf?p=geode-native.git;h=c46f42c ]

GEODE-8053: Fixes lingering StatsManager after Cache close. (#596)

While this was likely not a problem in C++ it was a big problem in .NET due to
the reliance on GC to destruct native objects. The StatsManager didn't stop
until it was destructed. This in turn stops the HostStatSampler. In .NET this
leaves the stats system trying to run on a closed but not destructed Cache
instance.

> Hang in legacy integration tests at shutdown
> --------------------------------------------
>
>                 Key: GEODE-8053
>                 URL: https://issues.apache.org/jira/browse/GEODE-8053
>             Project: Geode
>          Issue Type: Bug
>          Components: native client
>            Reporter: Blake Bender
>            Assignee: Jacob Barrett
>            Priority: Major
>             Fix For: 1.13.0
>
>
> When we switched the Windows build to release with symbols (RelWithDebInfo), we started hitting an apparent race condition at test shutdown that's blocking the release. What happens is, apparently:
>  * Test case starts running
>  * Cache is closed and deleted
>  * Stats thread wakes up, and tries to do something with stats
>  * Stats code calls a method on DataOutput that attempts to refer to this (now dead) cache object
>  * Test code crashes due to unhandled exception in stats code
>  * Test framework hangs, because it's waiting forever for child process to call ExitProcess explicitly (in a nutshell)
> We need to ensure that either a) the stats code never attempts to reference a dead cache, or b) when we hit the exception in the stats code, we catch it and exit gracefully.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)