Hi,
Yesterday evening and again this morning we've had two customers try
to upgrade their 32-bit Debian 10 VMs to Debian 11 and end up with
something that doesn't boot. This is because the Linux kernel
stopped supporting 32-bit Xen PV domains at version 5.9.
The quick workaround for those on Debian 10:
xen shell> virtmode pvh
xen shell> boot
We have talked a lot about this over on the "users" list over the
years, and for a while now the default at BitFolk has been 64-bit,
PVH mode guests, but we can't switch existing customers over to PVH
mode because it requires at least kernel version 3.19 and we don't
know what kernels you're running. So existing customers have been
left to switch on their own.
Switching to PVH mode will for now allow you to continue to run
32-bit VMs. However, aside from this, 32-bit Linux has been in
decline for some time and it's know to be less performant and less
secure than 64-bit. So the time has already passed where you should
be planning your switch to 64-bit.
== Just switching your kernel ==
Most of the advantage is to be gained by just switching the kernel,
so those running Debian could do that as Debian has good support for
this.
1. Upgrade to Debian 10 (buster)
2. Follow these instructions only up to and including the "Install a
kernel that supports both architectures in userland" step.
3. Connect to your Xen Shell
4. Shut down, boot, select the new amd64 kernel
xen shell> shutdown
xen shell> boot
If for any reason this does not work, just boot again and select
your previous i686 kernel again.
We suggest doing this in the Xen Shell so you can interact with
the boot process because the new amd64 kernel may not be listed
first in your bootloader.
5. Once satisfied that your amd64 kernel works you can remove the
i686 kernel packages.
Debian will take care of providing you with amd64 kernel updates in
future.
If you haven't already done so you should consider switching to PVH
mode now as well.
We do not recommend trying to fully cross-grade your operating
system to 64-bit unless you are an expert.
== Reinstall ==
You can do a reinstall in place yourself:
https://tools.bitfolk.com/wiki/Using_the_self-serve_net_installer
Don't forget to first switch your architecture to 64-bit and your
virtmode to PVH:
xen shelll> arch x86_64
xen shell> virtmode pvh
as these are the modern defaults.
We can also offer a new account free for two weeks for you to
install into and move your things over.
https://tools.bitfolk.com/wiki/Migrating_to_a_new_VPS
== PVH mode? ==
Which virtualisation mode you use is rather something we don't
expect customers to have to worry too much about, so new customers
have been in PVH mode for some time and haven't had to think about
it, but existing customers will need to make the change at some
point.
Anyone with a kernel that's 4.19 or newer should be able to switch
to it. Here's more info:
https://tools.bitfolk.com/wiki/PVH
Cheers,
Andy
--
https://bitfolk.com/ -- No-nonsense VPS hosting
Hi,
TL;DR: We turned off suspend+restore for everyone. We think it is
okay for you to re-enable it as long as you use kernel 4.2 or newer
(released 6 years ago), but can't tell what kernel you're running so
erred on the side of caution. We continue to use it for our own VMs.
More detail:
We've just opted you all out of suspend+restore because of the
filesystem corruption that afflicted 2 customer VMs during the
maintenance in August. There were 83 customer VMs that previously
had opted in.
While investigating that we did of course not do any suspend+restore
anyway. I am now satisfied that we know why it happened and under
what circumstances it should be safe to use it again, but as a
precaution we have opted everyone out of it so you can make your own
decisions.
A direct email has gone out to the main contact for each VM that had
previously opted in to this. That email contains far more detail. If
you think you had opted in to suspend+restore but don't see that
email please check your spam folders etc (and then mark it as "not
spam" if necessary!).
You can see the current setting (or opt back in) here:
https://panel.bitfolk.com/account/config/#prefs
You can read more about suspend+restore here:
https://tools.bitfolk.com/wiki/Suspend_and_restore
Thanks,
Andy
--
https://bitfolk.com/ -- No-nonsense VPS hosting
Hi,
We began receiving alerts at approximately 03:02Z today that host
"macallan" was unresponsive.
There was nothing interesting on its serial console. Its console
also did not respond. The out of band access to the BMC worked but
didn't show anything unusual. There were no hardware events logged.
In the face of a hard lock up all I could do was power cycle it.
All customer VMs were booted again by about 03:30Z.
I'll be keeping a close eye on this server. If this repeats then we
may have to move customers off of it at speed and with little
notice.
Apologies for the disruption this has caused.
Cheers,
Andy
--
https://bitfolk.com/ -- No-nonsense VPS hosting
