How to find out why an ESXi host is hanging?

Question

VMware

The story is this, inherited ESXi host 5.1 on desktop hardware, we decided to upgrade without looking to version 6.0. Everything went well, except for the fact that after the update, the drivers for the network card (RTL8169) disappeared. I googled that ESXi is slowly removing support for desktop hardware in favor of server hardware, and it looks like the driver has crashed. I found a manual on the internet, how to load drivers from 5.1 or from 5.5 Realtek 8169 and others to 6.0 via a USB flash drive. I did everything, everything worked. But after a couple of days, the host began to hang tightly. That is, until you reboot it by power. Before that (when it got there on version 5.1, the uptime was more than six months), here, after the update, I have not seen it for more than 3 days.

Because I am relatively new in this business, I ask for help, where and what to read and find in the logs, because of what this "subvison" is happening and what to do, but the host itself is not particularly loaded with guest OSes. And is it possible to "land" the host from outside?

Answer 1

I’ll leave this here because I often come across an article when googled
Total symptoms 1 in 1, only version 6.0
An article was found https://sysadminz.ru/index.php?topic=9765.0
as it turned out the problem can be even if you don’t use the raid
but update the driver the controller still did not, since the problem was with the backup server, it would be desirable to be without them.
but by typing in the command "dmesg | grep MSI" from the article Found an
error -13T08:02:22.697Z cpu4:33130)WARNING: LinPCI: LinuxPCI_EnableMSIX:862: 0000:00:14.0: Interrupt allocation failed with Not supported google found
https://kb.vmware.com/s/article/2106740
i.e. the whole thing is all the same in interrupts and the question is how to fix it
Disable this thing through the interface -
they disabled IO Virtualization - VT-d in the BIOS where)
While the flight is normal, if it does not help, we will update the controller