I wanted to see if anyone else has experienced some of the problems we are seeing and if/how you may have been able to resolve them.
My larger cluster consists of 4 SA 6000 appliances running 6.5R3HF5. On the average, I have about 2.5k-3k concurrent users spread across the appliances. CPU is typically at 15-20%
As for the problem itself; it seems the overall ability to navigate between menus on the appliances has progressivly degraded over the last year or so. I am at a point where bringing up the Active Users tab can take upwards of a minute and a half to two minutes. Accessing Web App profiles takes even longer than that. Granted, we have several hundred profiles so it could take a little time to load, but I cannot imagine that waiting 2-4 minutes is a normal behavior.
We have recently upgraded from 6.3R5 to 6.5R3HF5 but the slowness seems to persist. Just wanted to see if anyone else has run across this and what if anything you were able to do to resove this.
Best way to find is raise a case with JTAC and upload the system snapshot from each node.
Better to start in this way.
although Thovino's suggestion is a good one, you might look at what exactly you are syncing over the cluster (sessions, favorites, etc).
Also, keep in mind that any cluster traffic being passed between the nodes is sent over the internal port, you might check all parts of the internal LAN for heavy or otherwise abnormal load, port errors, etc.
Thanks for the advice.
I will likely open a JTAC case in a day or two. As for the sync settings, though we are in fact synching user sessions, neither the logs nor favorites are being syncronized across the cluster members. I have also noticed the issue exists independent of the number of user sessions.
I have a a lab cluster with the an identical configuration where I am experiencing the same problem. I cannot help but wonder whether the size of the configuration is a culprit. at my last count, i had 454 web resource profiles defined.
This was actually taken from a one month old config, I suspect we're closer to 480 by now. Based on the reaction to the number of profiles I believe it confirms my suspicion that it may be a bit much
As for the set up. These are base SA6000 appliances with no additional upgrades.
Other than managing the cluster via NSM, given the number of resource profiles we are dealing with, I'm not sure what if anything can really be done to improve the performance of web interface.
I've been discussing this with a colleague who has some experience with the 6xxx series and apparently, the number of profiles shouldn't be that much of a problem - not even for the admin interface.
Do you have any results from JTAC? I'm rather curious