Okay... you have the correct config file... cluster-service.xml

See the section that says PING Timeout---change the 2000 to 5000 (which is 5 seconds) 
(Bela may have other insight or preferences here... but this works for me).

PING Msgs is how J-Groups performs its status checks.  Upping the timeout to 5 seconds 
assumes that the multi-casts are actually delivered and allows your overloaded server 
time to run and reply to the pings.

In the FD section, change the timeout from 2500 to 4500, I noticed that your retries 
amount has already been bumped from 3 to 5... good.

Change the verify suspect from 3000 to 4750.

This is a good start and SHOULD handle many conditions.

You need to stabilize this mechanism before anything else can happen.  The J-Groups is 
fundamental to the cluster operating correctly along with the tree-cache and manny 
other services.

Once done, you will need to monitor your network ports (using TcpDump or similar tool) 
and watch the MCAST packets.  Watch the delay on the response packets.  If you see too 
much of a strectch between the response and request, then tweaking these numbers get's 
a bit trickier.  (A poor man's perf tool is PING.  You could just run a series of 
pings while your server is loaded and measure the response/stretch.  This is if'y at 
best since you don't have the Java Runtime and JBoss stack under you, but it does show 
machine/network latency).  If you see missing responses or pings from the cluster 
peers, then we need to determine why those are missing.  If you are using fast 
Ethernet ports and a lower end switch, chances are the multi-casts are being 
discarded.  Again, make your adjustments, and run the server.

When the load gets up there, try the ping test.  Also, if you can, capture a mintue or 
two with TCP dump and analyze the multi-cast packet flow and the time gaps between 
messages.

Once this is solid, we can move on to transactions, deadlocks, and rollbacks.


View the original post : 
http://www.jboss.org/index.html?module=bb&op=viewtopic&p=3840374#3840374

Reply to the post : 
http://www.jboss.org/index.html?module=bb&op=posting&mode=reply&p=3840374


-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 - 
digital self defense, top technical experts, no vendor pitches, 
unmatched networking opportunities. Visit www.blackhat.com
_______________________________________________
JBoss-user mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jboss-user

Reply via email to