Developer Only Discussion List

["Jeremy Blosser (emrys)" <jblosser-smgl AT firinn.org>]

On Oct 19, Adam Clark [adamsgl AT dhbit.ca] wrote:
> Now I've seen the suggestions and I think it would be a great idea to 
> have SMGL-purchased servers.  I can see this taking some time though, 
> for a few reasons - a) pieces need to be purchased and assembled before 
> the system can even have an install done, b) financially, SMGL doesn't 
> have things in place yet (as far as I understand), and I'm sure other 
> factors as to where would really be a good place to host the system 
> could take some time as well.

If we go with a dedicated server as we've been discussing, (a) is moot.
The server would be provided by the datacenter as soon as we requested it.

As for (b), as discussed I'm planning to front at least the first year.
I'll take donations to make back what I can but it's covered regardless.
A year is a decently long time... we'll either grow even more over the next
year and be in a place next year where more people can contribute, or we
won't grow and will decide to go back to free hosting somewhere else.  We
need something dedicated with an SLA, and if we can even try it for a year
that's something significant.

The location question (we'll call it (c)) was discussed in the other
thread; I didn't really hear any objections to sprocketdata.  It has a
*lot* going for it, more than any other place I can think of.

> I would like to make the argument though that the location where things 
> are hosted based on who we have 'available' currently is moot, or should 
> be a very low priority.  We've seen turnover at every level in the SMGL 
> organization, and it's entirely likely that whoever we may have admin an 
> SMGL server (or servers) at the physical level now, will probably not be 
> the same person or persons who's adminning it 6 months, or a year, or 3 
> years down the road.  I agree we should have an organization that's got 
> an SLA, can replace bits for us, etc.  But where we put servers based on 
> who's at what position in the SMGL structure should fall to a much lower 
> priority, imho.

I disagree that it's a low priority.  With all due respect, the physical
location of the servers is an overriding factor in the current downtime.
If the server was somewhere where more than one SMGLer could have reached
it, we could have gotten a rebuild moving much more quickly.  A server
that's going to stay up needs redundancy on critical components, and no
component is more critical during an outage than the human administrator(s)
with access to the server.

As for turnover, the three local admins that have volunteered in DFW would
be committed to maintaining this server whether we kept with SMGL or not.
Even if all three of us stopped using SMGL we would maintain our
responsibilities for that server until someone or somewhere else was found.
That's just how we do things.  Regardless, if anything the turnover is even
more reason to get it into a datacenter in an area with a higher
concentration of SMGLers, so any future handoffs can go easier.  In the
current scenario if you completely run out of time or decide to walk away
we're starting over from scratch.  I'm not suggesting you would do this,
but as my employers like to point out, anyone can get hit by a bus.  A
location with at least 3 people (4 counting dkowis) obviously requires more
people to walk away or get hit by a bus.  And even if we all go, if it's in
a datacenter other people have an option to show up and take it over.

> Anyhow, prior to seeing the -Admin postings, I had put together thoughts 
> on getting things back up... here's what I can offer right away:
> 
> I have a server available which has the following hardware:
> 
> - Tyan server board
> - 2 x 1.2Ghz P3 Tualatin cpu's
> - 1GB ECC RAM
> 
> It currently has 2 18GB SCSI drives in it, but space requirements are 
> much more, so was thinking of purchasing 2 200GB (or 160's, depending on 
> what's available today) IDE drives, using one of the SCSI drives as 
> boot/OS and the IDE drives as a software RAID 1'd setup.

Can you give us an idea of what our HD utilization was on the old server?
Also, any numbers you have on monthly transfer would be very useful.

> This system could be completely utilized for SMGL purposes, and could be 
> the basis on which other systems are slaved from.  It could be up and 
> running within a few days, too.
> 
> Whether this setup is useful or not, I do strongly agree that whatever 
> path we go down there should be off-site backups (and not just at one 
> location), and if we could do it, have servers that can be set very 
> quickly as master systems in the case that the primary server goes down.
> 
> At any rate, I'd appreciate any feedback, as I'd like to go pick up hard 
> drives tomorrow if this option is utilized at all.

I think we would all like to see a server back up at your location ASAP
with whatever latest backup we had so we can get in and assess the
situation and at least have some kind of bugzilla.  I would say you
shouldn't buy any hardware yet, assuming what you have is at least enough
to get the box operational.  I'm waiting on some final quotes from sprocket
on various hardware configurations; once I have this I'm pretty sure we're
ready to move forward on that option unless there's an overriding reason
not to.  If we do go that route we could use any money you would've spent
on hard drives toward the dedicated box, but of course that's entirely up
to you.

Attachment: pgpTFOwIMHoiM.pgp
Description: PGP signature