Memory usage of memory mapped files

General
1

Last week we looked at a Linux VM with 32 GB of RAM. Memory usage - according to Check-MKs “RAM + Swap overview” - was ridiculously low, about 0.9 of 32 GB. We reduced the RAM to 8 GB. Since then users were complaining about lousy performance.

Turned out: this VM runs an application that heavily uses mmap(2) to memory-map database files summing up to 90 GB. Of course performance dropped because now the VM was paging a lot more.

I can see the memory mapping using top: the database process has a VIRT allocation of 97.1g.

My question: is there any way Check-MK can stop me from falling into that trap again? How can I see how much memory is mapped to files and how active that memory is?

2

Have you looked at the other metrics of the Memory check?

This check records a ton of information.

1 Like
3

Yes I did. None of them reported more than 1 GB of usage of anything.

4

I think mmap()ed data is recorded in the “Mapped data” metric of that check which has not found its way into a graph. I do not know the reason of this decision here.

You can enable that graph by adding a new file $OMD_ROOT/local/share/check_mk/web/plugins/metrics/mem_mapped.py with this line as content:

check_metrics["check_mk-mem.linux"]["mapped"]["auto_graph"] = True
5

This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.