karaf-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jean-Baptiste Onofré (JIRA) <j...@apache.org>
Subject [jira] [Updated] (KARAF-4115) Hard coded timeout values for shell commands in Karaf and Cellar
Date Mon, 16 Nov 2015 16:04:11 GMT

     [ https://issues.apache.org/jira/browse/KARAF-4115?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Jean-Baptiste Onofré updated KARAF-4115:
----------------------------------------
    Affects Version/s: cellar-4.0.0
                       cellar-3.0.3

> Hard coded timeout values for shell commands in Karaf and Cellar
> ----------------------------------------------------------------
>
>                 Key: KARAF-4115
>                 URL: https://issues.apache.org/jira/browse/KARAF-4115
>             Project: Karaf
>          Issue Type: Bug
>          Components: cellar-core, cellar-shell, karaf-shell
>    Affects Versions: cellar-4.0.0, cellar-3.0.3, 3.0.5, 4.0.3
>            Reporter: Matej Ferencak
>            Assignee: Jean-Baptiste Onofré
>
> In Cellar, all commands seem to be set to timeout after 10s.
> This is causing us a lot of problems in situations when there are multiple Karaf Instances
deployed on multiple nodes (we usually deploy a total of 12-14 overall instances on 4 different
VMs), when Hazelcast tends to be really slow due to network traffic during cluster synchronization.
> In such a setup, running *cluster:group-set* can be very slow, and 50% of the time it
times out. Once this happens, the node is unuseable, i.e. it enters an undefined state where
it can not easily join or quit a cluster group any more.
> There are a couple of places in code where this can be found:
> org.apache.karaf.cellar.core.command.Command : timeout = 10000
> org.apache.karaf.cellar.core.event.EventDispatchTask : timeout = 10000
> (referencing Karaf 4.0.3 master)
> I can't find it right now, but a very similar thing seems to be implemented in Karaf
Shell, where I think I saw the value being bumped from 10s to a higher value of 30s at some
point in history (but still hard coded).
> Currently, we are working around this by re-implementing the *cluster:group-set* by invoking
Karaf Cellar services directly (ClusterManager & GroupManager). Since we implemented this,
we don't have problems with cluster set up any more.
> It would be great if this value could be somehow made configurable, or as the last resort,
increased.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message