[SOLVED] Rsync on LXC slow

[ropeguru]

So I have a file server setup in an LXC with 1TB of storage allocated backed by 1 3TB nvme drive. I have an offsite backup running on a Pi 4 with a 2TB USB3 drive attached and connectivity between the two using ZeroTIer.

I am testing some basic restores and transfer rates using rsync in to the lxc is giving me less than 5Mbps. Neither the lxc or the PI are showing any unusually high CPU usage. The lxc has 4 cores and 1GB ram assigned.

For testing purposes, I have a plex vm running on the same PVE server 2ith 4 cpu cores and 8GB ram, so I installed zerotier and tested the exact same rsync command and the restore is giving 100Mbps which is the max bandwidth from the remote end.

Why is there such a huge difference between the lxc and the vm in terms of throughput?

Edit: Posted here but I am not sure if this is a lxc vs. vm issue or a networking issue.

 

[ropeguru]

To add after more testing...

It appears that initially the transfer starts off at the same rate as the VM but then drops off to kiloByte speeds..

 

[ropeguru]

Update on findings with help from a fellow Redditor..

Looks like in the LXC as rsync runs, the OS is allowing the cahe to drop the free memory to zero and essentially causing a self denial of service. Running a VM with the same amount of RAM, the free memory drops to around 110KB but never goes any lower, so rsync continues to function as expected.

THe LXC is running a turnkey linux file server based on Debian 10 and the test VM is running Ubuntu Server.

 

[ropeguru]

I am going to close this one out and start a new thread in the correct forum now that I see it is an LXC memory issue.

 

[Matthias.]

Great! Please mark this thread as solved