Hypervisor hanging up - possible memory allocation issue

[borkode]

My server has recently been hanging at almost random times with the only way to sort it out is by force restarting the server - with the issue then happing again after a few weeks. From the looks of it, it seems to be a memory issue but what confuses me is that how the lxc's are able to cause the hypervisor itself to crash. Could someone please have a look at the logs and determine what could be the root cause?

logs:

https://pastebin.com/BwicZCnj
https://pastebin.com/KkD8FADf

any help would be appreciated

 

[Johannes S]

What do you expect? Lxcs ( and vms ) are basically Prozesses running on the hypervisor, so if a memory error least to a process crash why shouldn't it crash the host too? You can test for defect ram with memtest on the PVE install Media.

 

[pulipulichen]

Could you clarify what kind of storage you are using? Are you running on ZFS?

If you are using ZFS, it would be helpful to see your ARC statistics, as ZFS can sometimes be aggressive with memory allocation. Could you run arc_summary and check the following indicators for us?

ARC status: HEALTHY
Memory throttle count: 0

ARC size (current): 93.8 % 7.5 GiB
Target size (adaptive): 100.0 % 8.0 GiB
Min size (hard limit): 49.1 % 3.9 GiB
Max size (high water): 2:1 8.0 GiB

We are specifically looking to see if your Memory throttle count is above 0 or if the ARC size is hitting its limits, which could be related to the "hanging" and memory allocation issues you're experiencing.