Hi,
An update, since things are dragging on.
TL;DR:
Just at the point where everything seemed ready, one of the SSDs
appears to have developed a fault. I'm not comfortable pushing
things forward while there's uncertainty about that, so I'm going
through RMA with Intel over it before more progress can be made.
They've agreed without argument to a warranty replacement but I
don't at this point know how long that process will take. We are
looking at, I would guess, at least 2-3 weeks more delay.
Not Long Enough, Give Me More Info To Read:
The first new server has been used as a basis for upgrading
BitFolk's software to cope with the new version of Debian and of
Xen. The following features were rewritten to work here:
- Basic VM start/stop/console
- Rescue VM
- New OS installation
Still to work on:
- Creation/deletion/resize of block devices from Xen Shell
This is very rarely used feature, and can be handled in support
tickets so doesn't need to hold things up.
- Separation of charging for backup space from regular storage
This doesn't need to hold things up because the first customers
put on the new servers will not have any backup space configured.
Several BitFolk-internal VMs have been running on the new hardware
for a few weeks.
Great! Well…
About one month into the hardware testing and software development
phase, one of the Intel DC s3610 SSDs started throwing an ATA error
every couple of days.
I swapped the SSDs between slots and the problem followed the SSD,
so I am pretty confident that it's not an OS or motherboard problem.
The SSDs are really expensive (~£1k+VAT) and more importantly have a
huge lead time (20+ working days) so I am not keen to just buy
another.
A support case was opened with Intel who agreed pretty much
instantly to take it back for warranty replacement, although that
process has only just started and I do not yet know how long it will
take. If Intel tell me that it's going to take more time than just
buying another will take, then I will buy another in the mean time.
So, that is pretty much the last obstacle and it's a shame it's
happened but there is little I can do about that now. You can trust
that I am extremely motivated to get this sorted out, because not
only are we paying full price for the colo of a server that can't be
used for generating revenue, but we are also still paying for the
existing servers which customers need to be moved off of, which are
using triple the power (so triple the cost) for at best half the
performance!
Cheers,
Andy
--
http://bitfolk.com/ -- No-nonsense VPS hosting