From: Ferenc Gyurcsan (fgyurcsa@AVAILANT.COM)
Date: Fri Aug 16 2002 - 10:59:03 EDT
Besides, if I remember correctly, I read in a redbook that WebSphere is not
supported with any HA-like solution (such as HACMP). They said testing is
underway, so that may have changed in the meantime.
--Ferenc
-----Original Message-----
From: Aaron W Morris [mailto:Decep@NETSCAPE.NET]
Sent: Thursday, August 15, 2002 3:26 PM
To: aix-l@Princeton.EDU
Subject: Re: WebSphere and HACMP
The "HA" feature in WebSphere you are refering to is called Work Load
Management (WLM). WLM is more of a load-balancing solution rather than
fail-over solution, but fail-overs are an inherent feature.
Basically, you would have a WebSphere installation on 2 (or more) machines
with completely separate hostnames and IPs (using the same admin DB). The
application would be installed into a Server Group within Websphere and a
clone of the application could then be created on each server. The clones
are never "aware" of each other; the plug-in on the web server detects when
a clone (or server) is not responding and stops sending requests to it.
Then, every 60 seconds, the web server polls the dead clone until it detects
its presence (perhaps during a reboot) and starts sending requests to it.
HACMP would be a terrible waste of resources with WebSphere since you only
have 1/2 of your available resources available to you at any given time.
Plus, when a HACMP failover occurs, you could have up to 10 minutes when the
application would be unavailable. With WLM, the web server plug-in reacts
immediately, and there is never an apparent outage; the transition is
seamless even when the failed clone becomes available again.
My management wanted to use HACMP with WebSphere until I convinced of how
much extra work it would be as well as the benefits of WLM.
-- -Aaron W Morris If not fully satisfied with this email, please return unread portion for a full refund. Offer void in Utah. Other restrictions apply. "Christy Wu, T.F." <tfwu@TECHNOWORKS.NET> wrote: >Dear all, > >I have a 2-node HACMP cluster running a specific application X. The HA cluster is configured to hot standby mode and takes care of this specific application X which requires to bind to the hostname. Instead of installing Websphere to node A which is currently the production server, I'm required to install Websphere to node B for proction (as reuested by customer for certain restraints in his environment). Since Websphere by itself supports HA feature, I suppose if node B fails, Websphere in node A will somehow aware and take care of such failure. My concern here is that Websphere requires to bind to the hostname as well. How to resolve the hostname binding issure from both applications (Websphere and Application X), in fact it is a situration similar to mutual take over but only form application level of view. Are there any scripts or tricks to work around this solution? > >Best regards, >Christy > > __________________________________________________________________ Your favorite stores, helpful shopping tools and great gift ideas. Experience the convenience of buying online with Shop@Netscape! http://shopnow.netscape.com/ Get your own FREE, personal Netscape Mail account today at http://webmail.netscape.com/
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 22:16:09 EDT