Memory Leak? - Swap filling up gradually, resulting in getswapspace failures [Bug reported]

Not open for further replies.


May 8, 2016
After getting notifications about getswapspace failures on my FreeNAS box, I've been looking at its swap usage more closely and noticed that it fills up gradually over a couple of days until there's nothing left. Noticing a spike last night at 3 a.m., knowing that that's the time my VM runs a ~350MB backup to an NFS mounted dataset, I tried copying some data from the VM to the same NFS mounted dataset:

dd if=/dev/zero of=./largefileabc iflag=fullblock,count_bytes count=10G
dd if=/dev/urandom of=./largefileabc iflag=fullblock,count_bytes count=10G

The screenshots show spikes when copying 1GB, 10GB, and 20GB of zeros, then 1GB and 10GB of random. They show a definite effect on swap, but not linear with the size of the copied file at all:

MemLeak Copying Data NIC.PNG
MemLeak Copying Data swap.PNG

The following screenshot shows that with FN11.1-U5 and FN11.2 (since the last 'reset'), swap usage only ever decreases significantly with a reboot of the host.

FN11.1-11.2 Memory Leak.PNG

This has begun after I've started using VMs. I'd like to help properly debug this, but I have no idea what to do beyond these tests, which is why I ask here first, rather than submit a bug report straight away.

Can I do anything to investigate further, or should I go ahead and file this as a bug on the bug tracker?
Not open for further replies.