> Without digging them up I did a quick google on openbsd issues vultr.
> It pulled some things I saw before with 6.2 and timing, as well as
> issues with the base image, and other ones talking about a setting in
> KVM that was causing issues on certain servers.
I use Vultr, happily, but I _can_ confirm what you said. Randomly,
instances will 'freeze', and be unresponsive to the network, as well as
to the provided console. Looking at the settings, CPU has jumped up to
105% and plateaued, and that's it - you have to restart the instance.
I spoke to their support who were very helpful, telling me:
"This is an issue with OpenBSD on KVM/QEMU platforms that has yet to
be patched by the developers.
...
The issue is a bug in the kernel virtual machine (KVM) subsystem of
the Linux kernel; the flaw was introduced in version 4.10. It's
fixed in Linux kernel versions above 4.15.0."
Their solution is to use their 6.3 image, which automatically applies
the patch (on their end of the infrastructure), or set up your machine
your way, then open a ticket, letting them know the IP and they will
apply the patch manually and reboot the machine. I've done this, and
will monitor for any further erratic crashes.
Overall, their support has been good, and I can't speak to the speed of
their network, but my OpenBSD webservers have worked flawlessly (apart
from the above issue) on their infrastructure for many years.
Sacha
