Problems with Bandwidth

[Thomas15]

Hello!

I´m facing an issue with the bandwidth on some of our Proxmox VM´s. I tried to find a solution for this issue, but I am clueless what is causing this.

I had several VM´s running on an older node, everything was fine. Shortly after I did a Proxmox update on that node (I can not gurantee that this issue is 100% related to the update), VM´s started to have low bandwidth (sometimes 1Mbit Up and Down) while the bandwitdh of the hostsystem was absolutely fine (close to 10GBit)

The weird thing about this is that when you restarted the VM´s affected by this issue (not every VM had this issue) the bandwidth was fine for a short period of time immediately after the restart, usually for 2-10 minutes max. After that the speed was low again. The speed is also not static, so there is no cap somewhere which causes this issue.

After I established a new node, I decided to move all VM´s to the new system, and after migration this issue is still persistent for the VM´s affected. When creating a new VM on the node, the issue didn´t occur so far, this seems to be related to some of the exisiting VM´s only.

I am currently running "Proxmox Virtual Environment 4.2-2/725d76f0". Chaning the network cards (VirtIO, E1000 etc) of the affected VM´s doesn´t fix or change the issue in any way.

I would be glad if there is anybody who already faced and fixed the issue or if somebody has any idea what could cause this issue. I´ll glady share any additional information you require.

Thanks a lot in advance!

Greetings,
Thomas

PS: The old node was an OVH node (regarding to https://forum.proxmox.com/threads/vms-network-speed-slowing-down.28215/page-2) but the current one is hosted in Frankfurt, Germany.

 

[manu]

There could be a thousand factors reducing your network thoughput ( your hoster, a background backup job, other guests starting automatic updates ... )

* Can you provide some numbers about the throughput difference you notice ?
Like what would be the difference between wget http://path_to_my_iso_download.iso on host versus guest ?

* Also did you notice cpu usage spikes on the host when the network get slower ?

 

[Thomas15]

Hello manu,

thanks for your reply! I can asure you that there are no limitations on the hosters side. While speedtests on the affected VM´s show speeds down to like 1Mbit Up and Down, the speedtest on the hostsystem directly show results which are typcial for a 10GBit connection. (as long as the speedtest server can handle such a bandwidth) We tried this with usual speedtest and tested it with wget too, same results, very slow VM, lightning fast hostmachine.

There are also no problems with any load factors at all, no CPU, Memory or I/O issues. The bandwidth goes down minutes after the restart of the VPS and stays at such a low level. Without a restart, it never raises again.

Greetings!

 

[manu]

Does a LXC container exhibit the same behaviour ?
Do you have virtio drives in your guest VMs ?

 

[manu]

Hi
Have a look at this thread, it may be a problem of your hoster:
VMs Network speed slowing down

 

[Thomas15]

Hello manu,

yes, I refered to this thread in my first post already. I am on another hoster now, I were on OVH before. It turned out that the problem is an issue by the hoster, though. Since I had the problem with OVH and later even with the new hoster, I thought it could be a proxmox issue, but it obviously isn´t. It´s just bad luck that there is a pretty similar problem with the new hoster.

Thanks for your help!