Hello.
I have a server with Oracle Linux 7.9 installed (virtual server). I have a small database on the server (oracle). The server became inaccessible yesterday and today. When I turned the server off and on everything was fine. When I looked at the server through wmware the CPU and RAM usage was very high.
Properties
Utilization
8 CPU(s), 390 MHz used 64GB RAM
I could not detect anything via var/log/messages and dmesg.
There are no errors in the database.
The server is overloaded but I can't identify what it is. What would you recommend me to look for?
@soolan , what's the load if the VM is/not-running [with/without] the oracle instance running ?
what's top/htop showing ?
check using vmstat / pidstat / systat / sar .... all dependent upon being installed ( most should be)
expect other teammates to pitch in ....
I will add this information to the subject.
In what way was it "inaccessible"? Was trying to login on the console unresponsive?
If you restart the server then you can probably expect CPU and RAM utilisation to be very high for a while, especially if you have an Oracle dB instance to start up.
Remember that free CPU cycles are wasted CPU cycles and free RAM is wasted RAM. If the kernel memory manager is doing its job properly it is not going to swap out memory pages unless it has to.
What I'm saying is that the high CPU utilisation and high RAM utilisation may not be associated with the system becoming unresponsive until you can prove it.
I missed this detail. I wasn't pinging the server's ip
First, could you pls define "very high" and "overloaded"? We're always better off looking at numbers than at "feels". Did you run any commands to get usage info, or only VMWare's interface?
That said, can you log into the server via VMWare console? You mentioned dmesg and /var/log/messages, so I presume the server is accessible somehow.
In this case, pls review VMWare's network setting for this server - both the IP and "enabled" state (I remember getting servers out just by forgetting to enable some interface).
Make sure there are no more than one server with the same IP.
Also, check network interfaces status into the server with ip addr show. There should be one (and only one) interface with that IP and status UP.
Finally, make sure there is a valid route between the host you're running ip from and the server in question.
Read brendan gregg's book. Nobody can tell exactly about system performance issues.
I keep track of normal number of processes running on a mackine. It is very simple:
ps -ef |wc -l. When I think something is a miss I run the command again to see what is going on. I've had smart users log on to servers and use them as their work space. One guy managed to start 26k process useing some sort of recursive process. When we sshed into the server it took over 1 min to get the passwd prompt. When you have a loaded server take your time and have a look around before rebooting.
And there is a lot of things that can be done for tunning a machine for data base use. You could do a google search for something like "performance tunning your type of machine for your database" we would tune our Solaris servers for data base hosting. Since every thing is Oracle I'm sure they have some tec notes on system tunning for your dara base.