Hello fellow proxmox users / devs,
I recently upgraded to proxmox 5.2 (Which pushed the kernel version to 4.15) and faced a major slowdown on my network interfaces (on the host).
Instead of Gigabit(~110MB/s) I managed peaks of up to 6MB/s. I tried reconfiguring the cisco router, I swapped out every cable and even checked it by bridging the upgraded server via my other server (which was / is on kernel version 4.13)
I noticed that there is no issue with networking as long as the destination / source is in a subnet the server has in its own routing table (ip route).
As soon as it leaves that area, speed drops down to 6MB/s.
After 2 hours of testing I then tried downgrading the kernel and proxmox-ve which resolved the issue.
My server is using Broadcom eXtreme network cards (Driver bnx2x).
I found a few forum entries related to this issue, which I am not allowed to post due forum restrictions.
I will try out the proposed fix "sudo sysctl net.ipv4.tcp_congestion_control=cubic" this evening on a different server.
The same issue also appeared on my hetzner OOB server which was also successfully repaired by downgrading the kernel.
Anyone else faced the same issue (or similar)
I recently upgraded to proxmox 5.2 (Which pushed the kernel version to 4.15) and faced a major slowdown on my network interfaces (on the host).
Instead of Gigabit(~110MB/s) I managed peaks of up to 6MB/s. I tried reconfiguring the cisco router, I swapped out every cable and even checked it by bridging the upgraded server via my other server (which was / is on kernel version 4.13)
I noticed that there is no issue with networking as long as the destination / source is in a subnet the server has in its own routing table (ip route).
As soon as it leaves that area, speed drops down to 6MB/s.
After 2 hours of testing I then tried downgrading the kernel and proxmox-ve which resolved the issue.
My server is using Broadcom eXtreme network cards (Driver bnx2x).
I found a few forum entries related to this issue, which I am not allowed to post due forum restrictions.
I will try out the proposed fix "sudo sysctl net.ipv4.tcp_congestion_control=cubic" this evening on a different server.
The same issue also appeared on my hetzner OOB server which was also successfully repaired by downgrading the kernel.
Anyone else faced the same issue (or similar)
