[Linux-aus] VPAC etc

[Russell Coker]

On Tue, 7 Jun 2016 02:53:48 PM Noel Butler wrote:
> > As an aside the LUV server is now a VM on a Hetzner server.
> 
> Virtual services are more prone to outages then real hardware
> SANs are notoriously unreliable - we've seen them blamed for many virt
> env outages.

You can't directly compare things.  Linode has a virtual server offering that 
is quite reliable and gives good management options.  While a Linode VM will 
in some ways be less reliable than direct hardware access the options for 
management and recovery improve reliability and having Linode manage all the 
physical hardware offers benefits over running it all in-house.

I've had a lot of experience running servers for various organisations in 
various locations.  My experience is that Linode offers better uptime due to 
their management.

The LUV server is a VM on a hardware system that I have root on.  The Dom0 
deliberately has a fairly minimal configuration and most things are run on 
DomUs.  When there are problems with the LUV VM I can reboot it without risk 
as I can do whatever is necessary from the Dom0 to fix things.  For example I 
once had what should have been a routine upgrade crash and result in the LUV 
server being offline for 12 hours at VPAC until I could visit and fix it.  The 
exact same crash on the Hetzner VM would be fixed in minutes from Dom0.  The 
exact same crash on the Hetzner Dom0 would be fixed in less than an hour from 
the Hetzner recovery system.  On Linode it would also be fixed in less than an 
hour from their management console.

> real daemons running on real disks and not from a big file are far far
> far more efficient.

That depends on what you are doing.  Linode uses SSD and will outperform any 
real disk.  Hetzner servers have the option of using SSD, in the affordable 
options of Hetzner servers you can have 2*512G SSD or 2*4TB disk.

Linode had some disk performance issues before they moved to SSD, but with SSD 
performance is always great.

> You know your hardware, so if it fails its your fault.
> You know your hardware so you know what it is capable of

There are economies of scale in running hundreds or thousands of identical 
systems.

> As for external hosting with the above named services, multiple ip
> ranges of them, especially hetzner and linode are dnsbl'd.

I haven't had problems with that in the past.  But really no matter what IP 
ranges you use there are always possible problems.

On Tue, 7 Jun 2016 03:07:17 PM Joshua Hesketh wrote:
> > Whilst Linux Australia had undertaken all possible means to ensure the
> > equipment was identified as belonging to the organisation, ownership
> > was mistakenly attributed to a Victorian linux users group, who also
> > had equipment hosted in the room. The room was decommissioned at 9
> > weeks ahead of the last shutdown date given, which meant that
> > migration works had not yet been completed.

In retrospect it would have been a better idea for LA to work with people from 
LUV in arranging these things.  VPAC had 3 active LUV members on staff who 
might have been available to help out.  Also it it might be a good idea to 
contact locals wheneven LA is doing something in a particular region, I'm sure 
that there are other LUV members who could have helped.

-- 
My Main Blog         http://etbe.coker.com.au/
My Documents Blog    http://doc.coker.com.au/