[
https://issues.apache.org/jira/browse/UIMA-4829?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15269479#comment-15269479
]
Lou DeGenaro edited comment on UIMA-4829 at 5/4/16 11:06 AM:
-------------------------------------------------------------
When SM sets the state of a service, it now considers whether the pinger is
stale.
If the current state is Available and the pinger is stale, then force the state
to become Waiting. This will block allocation of resources for those
Jobs/Services that have a dependency on the now Waiting service. Once the
pinger becomes un-stale, the SM will change the service's state back to
Available.
Pinger staleness occurs when the last update time from the pinger exceeds
ducc.sm.meta.ping.stability * ducc.sm.meta.ping.rate, which are already
existing ducc.properties variables.
How many consecutive missed pings are required to mark a service unresponsive
ducc.sm.meta.ping.stability = 10
How often services are monitored to insure they're responding, in milliseconds
ducc.sm.meta.ping.rate = 60000
was (Author: lou.degenaro):
When SM sets the state of a service, it now considers whether the pinger is
stale.
If the current state is Available and the pinger is stale, then force the state
to become Waiting. This will block allocation of resources for those
Jobs/Services that have a dependency on the now Waiting service. Once the
pinger becomes un-stale, the SM will change the service's state back to
Available.
Pinger staleness occurs when the last update time from the pinger exceeds
ducc.sm.meta.ping.stability * ducc.sm.meta.ping.rate, which are already
existing ducc.properties variables.
# How many consecutive missed pings are required to mark a service unresponsive
ducc.sm.meta.ping.stability = 10
# How often services are monitored to insure they're responding, in milliseconds
ducc.sm.meta.ping.rate = 60000
> When a DUCC pinger hangs it is reported as Available
> ----------------------------------------------------
>
> Key: UIMA-4829
> URL: https://issues.apache.org/jira/browse/UIMA-4829
> Project: UIMA
> Issue Type: Bug
> Components: DUCC
> Reporter: Burn Lewis
> Assignee: Lou DeGenaro
> Priority: Minor
> Fix For: 2.1.0-Ducc
>
>
> The DUCC pinger can hang and the main evidence is the stale ping statistics.
> The SM should mark it as unavailable and indicate that pinger is active but
> not responding.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
