Today Nagios Monitor discovers an Oracle database server is heavily swap-out, causing some Nagios monitoring item timeout alarms
Viewing the server, the swap space is set to 32G, 19G is used, the usage is very high, and the vmstat shows that Si is severely
This server physical memory 32g,sga set 20G, the number of connections is not high, even if set improperly, should not have such a serious swap
Cat/proc/meminfo
Hugepage is found on this server, but the status is free
We know that Hugepage set, even if not use it, the occupied memory space can not be used by other processes, and, hugepage is the pin in the physical memory space, will not be swap, it means 32G physical memory, actually only 12G usable, no wonder so high swap
So why does this problem occur? Why didn't this problem occur before?
The reason for this is that the server was restarted 2 months ago before the engineer ran the Ulimit-l unlimited before starting Oracle, and the engineer restarted 2 months ago did not know that the hugepage was turned on and restarted Oracle directly in standard mode. Unable to use hugepage due to an issue with operating system limit configuration
The workaround is simple
1. Before starting Oracle, perform Ulimit-l Unlimited
2. Modify/etc/security/limits.conf
Increase:
Oracle Soft Memlock-1
Oracle Hard Memlock-1
[@more @]
The server on which the Oracle database is on swap critical