You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@geode.apache.org by "Patrick Rhomberg (JIRA)" <ji...@apache.org> on 2018/02/16 17:41:00 UTC
[jira] [Updated] (GEODE-4643) Shutdown command should (optionally)
wait for shutdown
[ https://issues.apache.org/jira/browse/GEODE-4643?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Patrick Rhomberg updated GEODE-4643:
------------------------------------
Description:
Shutdown of a member is asynchronous. The {{shutdown}} command will trigger shutdown, but not guarantee it.
Beyond the obvious benefit of guaranteeing cluster state, this has also proven problematic in the {{runAll}} task of the {{geode-examples}} repository, as subsequent tests can begin before the previous test has finished tearing down.
-----
Update:
The addition of wait tasks in GEODE-4666 have not resolved the CI failures in the {{geode-examples}} gradle task {{runAll}}. This suggests that the {{shutdown}} command can fail to actually stop all members. Due to an error in the initial commit of GEODE-4666, the 10 second timeout was ignored, resulting in a ~45 minute timeout in Travis. As such, is it reasonable to believe that shutdown either failed entirely or never executed on at least one member.
was:
Shutdown of a member is asynchronous. The {{shutdown}} command will trigger shutdown, but not guarantee it.
Beyond the obvious benefit of guaranteeing cluster state, this has also proven problematic in the {{runAll}} task of the {{geode-examples}} repository, as subsequent tests can begin before the previous test has finished tearing down.
> Shutdown command should (optionally) wait for shutdown
> ------------------------------------------------------
>
> Key: GEODE-4643
> URL: https://issues.apache.org/jira/browse/GEODE-4643
> Project: Geode
> Issue Type: Bug
> Components: gfsh, management
> Reporter: Patrick Rhomberg
> Priority: Major
>
> Shutdown of a member is asynchronous. The {{shutdown}} command will trigger shutdown, but not guarantee it.
> Beyond the obvious benefit of guaranteeing cluster state, this has also proven problematic in the {{runAll}} task of the {{geode-examples}} repository, as subsequent tests can begin before the previous test has finished tearing down.
> -----
> Update:
> The addition of wait tasks in GEODE-4666 have not resolved the CI failures in the {{geode-examples}} gradle task {{runAll}}. This suggests that the {{shutdown}} command can fail to actually stop all members. Due to an error in the initial commit of GEODE-4666, the 10 second timeout was ignored, resulting in a ~45 minute timeout in Travis. As such, is it reasonable to believe that shutdown either failed entirely or never executed on at least one member.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)