[SAC] spatialref.org

[Alex Mandel]

There is overhead to a 100% separate VM (Hardware and People). Leaving a 
system for 7+ years without security patches is also a recipe for it 
becoming subject to a botnet (recent trend is to hijack servers instead 
of desktops). I won't rule it out and suggest that if that's what we 
want someone needs to bug OSUOSL about the backup server we wanted (I've 
gotten no response lately) so that we free up space on osgeo4 for more 
VMs, or we approach the new EDU labs about load balance hosting (my lab 
would probably offer). We might also consider using CloudFlare at the 
free level to get world wide DOS protection and CDN(side benefit).

In terms of isolation, python virtualenv is a decent compromise, and 
tends to only break when python itself if upgraded. I had already 
isolated the wsgi component to gunicorn a few years ago, so that apache 
doesn't really effect it.

Such is the life of the projects VM. If we do feel we need more 
stability there are other VMs that see less changes, host fewer sites we 
could shift to.

FYI, from what I can tell, spatialreference.org is now Django 1.4.5 
compatible. I spent about 1 hour making that happen which isn't terrible.

Thanks,
Alex

On 06/11/2013 09:44 AM, Basques, Bob (CI-StPaul) wrote:
> All,
>
> I'm willing to help on this, but need to share some recent experiences as a disclaimer (of sorts) . . .
>
> Well first, since it's a service (most importantly) I would think putting it on it's own VM/hardware (well actually Hardware/VM is my preferred order, read on) would be the first thing to do.  Services should not go down or get broken because of infrastructure changes to a service (in my mind).  I tend to opt for leaving a service alone once it's up and running, it can be augmented, but any sort of major redo suggests setting up the service all over again, which sounds like where the SR.org site is at in it's lifecycle.
>
> I'm doing something similar here with a 7 year old system as a matter of fact, it's run just fine over the years, in a 24/7 mode, going down a total of three times while on it's own hardware, all three times being network related and beyond my control.  I'm much more comfortable with using a new piece of hardware instead of being at the mercy of a VM farm, which seems to go up/down almost weekly for maintenance of this or that, and which I have little control over.
>
> In the interests of stability, I tend to use dedicated hardware for (each) service, does it cost more (I would argue no, based on the longevity of a service).
>
> Ok, that's off my chest, so do you still want me to help  :c)
>
> Can someone shoot me the last couple of years of usage logs, or even a usage summary of the site?  I may have some options for homing on dedicated space if that's of interest.
>
> Bobb
>
>
> -----Original Message-----
> From: sac-bounces at lists.osgeo.org [mailto:sac-bounces at lists.osgeo.org] On Behalf Of Howard Butler
> Sent: Tuesday, June 11, 2013 11:29 AM
> To: System Administration Committee Discussion/OSGeo
> Subject: Re: [SAC] spatialref.org
>
>
> On Jun 11, 2013, at 11:25 AM, Frank Warmerdam <warmerdam at pobox.com> wrote:
>
>> Bobb,
>>
>> It is now a subproject of the OSGeo Metacrs project and it hosted on OSGeo hardware.  I think Howard Butler, Chris Schmidt, Dane Springmeyer and Josh Livni were the original authors (forgive me if I'm wrong).
>
> Yup, I transferred the domain name to OSGeo a few years ago. It's there for anyone willing to step in it, but as I said, it needs more than a new coat of paint.
>
> Howard
>
> _______________________________________________
> Sac mailing list
> Sac at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/sac
>
>
> _______________________________________________
> Sac mailing list
> Sac at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/sac
>