[SOLVED] VM crashes regularly

[mijo112]

Hi, my HomeAssistantOS VM (ID:100) crashes regularly. I can't see any reason in the log.
Enclosed is the log, what exactly does it say? The host is a Lenovo m720q.
Due to the hardware hang in the log, I found the following workaound in the forum. If the workouaund is entered, the crashes are more frequent, but even without the option the VM crashes.

Workaround:

ethtool -K eno1 tso off gso off

Network card hang-ups.:

Feb 09 16:09:11 proxmox kernel: e1000e 0000:00:1f.6 eno1: Detected Hardware Unit Hang:
  TDH                  <8e>
  TDT                  <aa>
  next_to_use          <aa>
  next_to_clean        <8e>
buffer_info[next_to_clean]:
  time_stamp           <10082b2c0>
  next_to_watch        <8f>
  jiffies              <10082bbc0>
  next_to_watch.status <0>
MAC Status             <40080083>
PHY Status             <796d>
PHY 1000BASE-T Status  <3800>
PHY Extended Status    <3000>
PCI Status             <10>

Crash log in the Appendix.
Begin:

qemu-system-x86 invoked oom-killer: gfp_mask=0x140cca(GFP_HIGHUSER_MOVABLE|__GFP_COMP), order=0, oom_score_adj=0

 

[PwrBank]

Out of curiosity, why use the e1000 virtual NIC instead of a VirtIO?

 

[mijo112]

All CT and VM use VirtIO, only the host has an Intel network card.

Out of curiosity, why use the e1000 virtual NIC instead of a VirtIO?

 

[fiona]

Hi,

Crash log in the Appendix.
Begin:

qemu-system-x86 invoked oom-killer: gfp_mask=0x140cca(GFP_HIGHUSER_MOVABLE|__GFP_COMP), order=0, oom_score_adj=0

oom stands for out of memory. So your host/Linux does not have enough RAM at some point and needs to reclaim it (which it does by killing the VM).

Regarding the hardware unit hang, that is a known hardware/firmware issue, for making the workaround automatic see: https://forum.proxmox.com/threads/e1000-driver-hang.58284/page-7#post-368615

 

[mijo112]

Hi, I wanted to give some feedback.
It was due to too little RAM. After the upgrade it is now running stable.
With the ethtool workround there were more crashes, because the workaound has an extended RAM consumption.
It would just be nice if there were warnings in the UI when there is too little RAM, as with VMware, for example.