I would like to configure the syslog.conf to have a good monitoring information about my system.
do you have any idea about best configuration from your experience in your Data Centers
BR,
Usually, the defaults are fine until you start debugging a problem. You can turn everything on for a new file for a limited time, and see what you do not want, turn that off and try again. Syslog is used by any application that thinks it is appropriate, but usually daemons of some sort, so it can be hard to generalize. Some tags are explicitly for detailed debug messages.
ok thanks alot, do you know a free tool to monitor the servers
Hi Maxim42,
There is a forum here that covers monitoring tools, my preference would be xymon - but there are lots of them.
Check out the infrastructure monitoring forum.
Regards
Dave
thanks alot
Often, nothing less than some sort of "end to end test of each server", running on a dedicated testing host, will assure you that they are all alive. This is one reason I dislike "warm spare" architecture, where a host sits unused until a failover. It might be dead or the failover may be a failure. I am for divided load and application specific server testing. With the right architecture, the load can shift pretty instantly to the remaining server(s) when one dies, give or take any services in process there.
Like I said in one memorable customer meeting on reliability, "Sometimes the system is not down but just running very slowly." They asked "How can you tell if it is too slow>" I said, many times in the next hour, "A loud banging on the computer room door!" Testing should test not only function but latency/response-time. Good monitoring shows you when things are running slow and backing up, so you do not accidentally kill a live server trying to service it. There are many ways to make a server temporarily slow, like paging every page out. Even without thrashing, one good sweep can make for a million page faults as other apps reestablish themselves in RAM.