Hi,
I am going to cluster our notes apps servers. We have lot of dbs with scheduled notes agents where server name is mentioned to run on.
If I cluster these dbs, will agents failover to the clustered server ? any ideas, pls let me know
-Archana
Subject: Failover of Agents in Cluster
If you want these agents to run on the servers in the clustered environment, you need to choose “run on any server”;
If you do not want them to run again, then choose only the production server to run them on and they will not run on the clustered server.
Are you going to be using the cluster as a Disastor recovery site, or do you want an actual failover. I ask this because I just spent 2 months setting up my clusters for disastor recovery so that everything replicates but failover happens only when I want it to.
Usres get confused when they start getting more than 1 icon on their workspace and there is not immediate transfer of data.
Subject: RE: Failover of Agents in Cluster
Thank you for your reply. I am setting up mail cluster for load balance and fail-over. Currently I have agents set up to run on primary server only. As per your suggestion then I have to go each and every agent and change it to “run on -any- server” Is that right ?
Also please let me know any of your other experiences with clustering…
thanks
archana
Subject: RE: Failover of Agents in Cluster
YOu need to be careful with this setting, as it will run on EVERY server. You need to add more logic to your agent to prevent unwanted execution and reprocessing of documents. This is a very complex issue.
Before delivery mail agents, support cluster failover natively in Domino. (See Agent FAQ, “Out of inbox” for a longer discussion).
For other agents, there is no built-in logic. There was an article in a magazine View where one of the Lotus developers shared how cluster support was implemented in Domino.Doc.
Subject: RE: Failover of Agents in Cluster
You have to be careful when setting up agents to run on both the primary server and the clustered server. You primarily want to run agents on one server and replicate the results to the other server.
Running agents on 2 servers against the same db could produce some unexpected results like:
-
<replication/save conflicts>
-
If an agent runs to notify someone of an event, then you may notify that peson twice.
-
Data updated by time, may be out of sync.
You really need to think it through before you act on how you want your agents running and your cluster setup.
I have worked in both environments and have been exposed to some Yucky clustering issues. The way I have it setup today seems to work very effectively with the least impact to my users.
That is clustering with:
The SERVER_AVAILABILITY_THRESHOLD=100 & the SERVER_RESTRICTED=2 on the clustered server.
Both of these flags can be toggled in the event a server goes down without restarting the server. The users do not get automatically bumped around in the event a server just slows down for 30 seconds. Replication is immediate because the clustering is turned on.
Do you really want your users bounced between icons on their desktop? Help desk calls go up because users will never grasp the concept of clustering and the difference between the 1-5 minute interval between updates. All they know is that they keep getting pushed back and forth to different servers. If mail is running on the web only, then clustering wouldn’t matter but sounds like you are in the client.
Hope all this helps…
Subject: RE: Failover of Agents in Cluster
Julie:
Are you talking about this article?
Failover Support for Background Agents on Clustered Domino Servers — A Solution for Your Lotus Workflow and Domino Applications
by Luciano Resende
I have the same problem as Archana
JYR
Quebec,Canada
Subject: RE: Failover of Agents in Cluster
Yes! Thank you for posting the link!
Subject: Prioritize Servers in Agents
Change setting to run on any server. Then, to accomplish this in a clustered environment using Lotusscript agents (or java), create an array of server names with the primary server listed first. Then the first thing the agent should do is to check the list, to see if the current server is the primary server. If so, then run the agent. If it tries to run on the other servers, it should check to see if the primary server is available (up and running), if it is, stop processing, else check the next server in the list.
This would create a true failover and avoid replication conflicts.
Subject: RE: Prioritize Servers in Agents
One scenario that might break this is a simple network outage between the 2 servers. This could occur for example, if the primary and cluster server were located at different sites (Prod & DR Site…)
Subject: RE: Failover of Agents in Cluster
Hi JY,
I think this article will be really useful to resolve this issue. Do you have this article in PDF or in some other format.
Can you please share it as I don’t have 2002 view journals.
thanks
archana
email :asonsurkar@yahoo.com
Subject: RE: Failover of Agents in Cluster
You can order one free issue from the magazine. Asking someone else to send you a copy would be violating copyright.