PVE with Linux 4.15 has a network problem

[octset]

hello:

I have server in online.net

I reninstall a new debian9 and install PVE on it

After installation I reboot and test

I found that my network speed has been reduced from 1000M to 100M.




I was very surprised, after I re-installed, completely pure system, I found that the speed of the network dropped after installing PVE

Finally, I set the boot menu in IPMI and selected the Debian kernel.




Now my internet speed is normal.





What caused this? please help me


Thank you very much！！！

 

[octset]

I am pretty sure that two weeks ago, I installed it in the same way, the PVE kernel is available, is it updated recently?

 

[Stoiko Ivanov]

Could you please post the output of `ethtool <your-interface>` , as well as the output of `lspci -nnk`? (when running the PVE kernel)
What NIC do you use?
Do you see anything in the output of `dmesg` or the journal ?

 

[octset]

Could you please post the output of `ethtool <your-interface>` , as well as the output of `lspci -nnk`? (when running the PVE kernel)
What NIC do you use?
Do you see anything in the output of `dmesg` or the journal ?

Settings for eno1:
Supported ports: [ TP ]
Supported link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
1000baseT/Full
Supported pause frame use: Symmetric
Supports auto-negotiation: Yes
Advertised link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
1000baseT/Full
Advertised pause frame use: Symmetric
Advertised auto-negotiation: Yes
Speed: 1000Mb/s
Duplex: Full
Port: Twisted Pair
PHYAD: 0
Transceiver: internal
Auto-negotiation: on
MDI-X: on (auto)
Supports Wake-on: pumbg
Wake-on: g
Current message level: 0x00000007 (7)
drv probe link
Link detected: yes

00:14.0 Ethernet controller [0200]: Intel Corporation Ethernet Connection I354 [8086:1f41] (rev 03)
Subsystem: Super Micro Computer Inc Ethernet Connection I354 [15d9:1f41]
Kernel driver in use: igb
Kernel modules: igb
00:14.1 Ethernet controller [0200]: Intel Corporation Ethernet Connection I354 [8086:1f41] (rev 03)
Subsystem: Super Micro Computer Inc Ethernet Connection I354 [15d9:1f41]
Kernel driver in use: igb
Kernel modules: igb

 

[octset]

I only use one of the them , the other one is not connected.

 

[Stoiko Ivanov]

Speed: 1000Mb/s

At least on the link-level you're still running with 1G

While this would be a huge impact - it could be the result of the Spectre mitigation - maybe try to disable them (keep in mind that there are quite some security implications!) and see whether you get your speed back up.

But before maybe check your logs and journal for something related to the drop in speed

 

[octset]

At least on the link-level you're still running with 1G

While this would be a huge impact - it could be the result of the Spectre mitigation - maybe try to disable them (keep in mind that there are quite some security implications!) and see whether you get your speed back up.

But before maybe check your logs and journal for something related to the drop in speed

So I can only use the 4.9 kernel now?

But I am pretty sure that a few weeks ago, the same installation method, the same steps, did not appear such a problem

What is the update during this time?

 

[octset]

I install PVE on Hetzner also found this problem

 

[Stoiko Ivanov]

You could try disabling the various offloading features provided by the NIC:
Check which offloading features are available, and `on`, by running

ethtool -k eno1

Disabling them works like this for example:

ethtool -K eno1 gro off tso off rxvlan off

The features and concrete syntax is mostly available in the ethtool manpage.

 

[anotherGregg]

I was experiencing problems with e1000 driver and

4.15.18-7-pve #1 SMP PVE 4.15.18-27 (Wed, 10 Oct 2018 10:50:11 +0200) x86_64 GNU/Linux

the symptoms are:

• guest os temporary freezes
  
• syslog on guest vm may have entries like

kernel:[136529.224343] NMI watchdog: BUG: soft lockup - CPU#0 stuck for 31s! [cron:6735]

• syslog on host has entries like

kernel: [195137.204691] e1000e 0000:00:1f.6 enp0s31f6: Detected Hardware Unit Hang:

and, finally, host freezes and only reset button can bring it to life.

using following advice

You could try disabling the various offloading features provided by the NIC:

and stackoverflow answers, looks like disabling tcp checksum checks solves my case

ethtool -K <YOUR_NIC_NAME_HERE> tx off rx off

Did not wanted to start new thread, hope this will be helpful