Something is eating my RAM

[Andreas Pflug]

In a cluster of PVE 7.0 machines, I observe that machines have much less free memory than expected. I evacuated one machine, and rebooted.
After the reboot, free shows 8GB used of 32GB total. 24h later, the RAM used has increased to 12GB. I can't locate any responsible process for this: htop lists linstor-satellite with 0.9% memory usage and corosync with 0.6%.

The machine still has no VMs on it, only drbd 9 and gluster running to serve the pool.

Any idea what might eat the memory, or how I can locate the consumer?

 

[t.lamprecht]

Hi,

are you sure it's not just Linux using the available memory for the page cache/buffers, which can be made available very fast once there's actual demand for that memory?

Can you please post the output of the free command?

 

[Andreas Pflug]

total        used        free      shared  buff/cache   available
Mem:        32790452    13066656    13825864       62860     5897932    19125360

The Memory Usage graph of that node tells the same story. Excerpt from /proc/meminfo:

MemTotal:       32790452 kB
MemFree:        13826080 kB
MemAvailable:   19125576 kB
Buffers:          127636 kB
Cached:          5450004 kB
Active:           351244 kB
Inactive:        7691464 kB
Active(anon):       2700 kB
Inactive(anon):  2495376 kB
Active(file):     348544 kB
Inactive(file):  5196088 kB

 

[patch]

32GB total. 24h later, the RAM used has increased to 12GB.

ZFS using 50% of available ram??
If you have ZFS storage, what does arc_summary show?

 

[Andreas Pflug]

ZFS using 50% of available ram??
If you have ZFS storage, what does arc_summary show?

no ZFS. FWIW, just checked on a machine with ZFS, can't see the consumed ARC memory in /proc/meminfo or htop either. So apparently kernel modules may consume memory that don't show up where I look.