enabling update.config in a cluster

["Earle, Erik" <Er...@disney.com>]

Hi gang!   If update.config is enabled on a cluster with a URI to fetch periodically, does each node do the fetch or is there a mechanism so only one node in the cluster is making the request?

RE: enabling update.config in a cluster

["Earle, Erik" <Er...@disney.com>]

I'm looking at strategies for this. There needs to be some way for each member of a cluster to know if it should make the request or let another member be the host that makes the request.  Are there cluster level stats (like member uptime or member join order) available on all the hosts?  ...or is there an obvious (to those that know the apis well) way to do 'leader election' or distributed locking in a TS cluster that I can get a nudge toward?


 

-----Original Message-----
From: ming.zym@gmail.com [mailto:ming.zym@gmail.com] 
Sent: Wednesday, September 26, 2012 9:58 PM
To: users@trafficserver.apache.org
Subject: Re: enabling update.config in a cluster

the answer is NO for the current codes, the scheduled update is on the
transform hooking layer, and the codes is not cluster aware right now.
we may make this a feature request, technically we can implement it.

file the bug and submit the patch if you like :D

thanks

在 2012-09-26三的 10:14 -0700，Earle, Erik写道：
> Hi gang!   If update.config is enabled on a cluster with a URI to
> fetch periodically, does each node do the fetch or is there a
> mechanism so only one node in the cluster is making the request?

Re: enabling update.config in a cluster

["ming.zym@gmail.com" <mi...@gmail.com>]

the answer is NO for the current codes, the scheduled update is on the
transform hooking layer, and the codes is not cluster aware right now.
we may make this a feature request, technically we can implement it.

file the bug and submit the patch if you like :D

thanks

在 2012-09-26三的 10:14 -0700，Earle, Erik写道：
> Hi gang!   If update.config is enabled on a cluster with a URI to
> fetch periodically, does each node do the fetch or is there a
> mechanism so only one node in the cluster is making the request?