+1 Thanks for bringing this up Raj. Yes we should do the changes.
BTW hope you are feeling well now :-)

Thanks

On Thu, Feb 12, 2015 at 9:05 AM, Rajkumar Rajaratnam <[email protected]>
wrote:

> Hi Devs,
>
> Sometimes back, we had different cluster monitors for VMs vs Kubernetes
> and Service vs LB clusters. So we kept different cluster monitoring
> intervals and member expiry timeouts. We made these configurable via
> autoscaler.xml like below;
>
>             <member>
>                 <vm>
>                     <!-- this is the maximum time(ms) a vm member can be
> in pending member state -->
>
> <pendingMemberExpiryTimeout>900000</pendingMemberExpiryTimeout>
>                     <!-- this is the maximum time(ms) a vm member can be
> in obsoleted member state -->
>
> <obsoletedMemberExpiryTimeout>86400000</obsoletedMemberExpiryTimeout>
>                 </vm>
>                 <container>
>                     <!-- this is the maximum time(ms) a container member
> can be in pending member state -->
>
> <pendingMemberExpiryTimeout>900000</pendingMemberExpiryTimeout>
>                     <!-- this is the maximum time(ms) a container member
> can be in obsoleted member state -->
>
> <obsoletedMemberExpiryTimeout>3600000</obsoletedMemberExpiryTimeout>
>                 </container>
>             </member>
>             <!-- cluster monitoring interval -->
>             <monitorInterval>
>                 <vm>
>                      <!-- VM Service cluster monitoring interval(ms) -->
>                      <service>90000</service>
>                      <!-- VM LB cluster monitoring interval(ms) -->
>                      <lb>90000</lb>
>                 </vm>
>                 <kubernetes>
>                      <!-- Kubernetes Service cluster monitoring
> interval(ms) -->
>                      <service>60000</service>
>                 </kubernetes>
>             </monitorInterval>
>
>
> Since we are having a single cluster monitor now, most of these parameters
> are not used and should be removed from autoscaler.xml. And there are no
> difference between VM, Kubernetes, Service and LB clusters. Hence, we
> should change it to following format.
>
>             <member>
>                     <!-- this is the maximum time(ms) a member can be in
> pending member state -->
>
> <pendingMemberExpiryTimeout>900000</pendingMemberExpiryTimeout>
>                     <!-- this is the maximum time(ms) a member can be in
> obsoleted member state -->
>
> <obsoletedMemberExpiryTimeout>86400000</obsoletedMemberExpiryTimeout>
>                     <!-- this is the maximum time(ms) a member can be in
> pending termination member state -->
>                     <pendingTerminationMemberExpiryTimeout>1800000</
> pendingTerminationMemberExpiryTimeout>
>            </member>
>
>            <cluster>
>                     <!-- cluster monitoring interval (ms) -->
>                     <monitorInterval>90000</monitorInterval>
>             </cluster>
>
> We should also update the code-base to read the correct parameters
> accordingly.
>
> wdyt?
>
> I will do the necessary changes.
>
> Thanks.
>
> --
> Rajkumar Rajaratnam
> Committer & PMC Member, Apache Stratos
> Software Engineer, WSO2
>
> Mobile : +94777568639
> Blog : rajkumarr.com
>



-- 
Imesh Gunaratne

Technical Lead, WSO2
Committer & PMC Member, Apache Stratos

Reply via email to