Re: [Slony1-general] Migrating From gBorg

Moving to -www, since this is more than just a Slony issue.

On Tue, Sep 19, 2006 at 12:21:33PM -0300, User Marc wrote:
> On Tue, 19 Sep 2006, Dave Page wrote:
>
> > No, it's common sense. Marc's day job is running his servers which he
> > knows inside out. Mine is running Vale Housing's servers. I'll back up
> > Marc as far as the postgresql.org resources are concerned but neither I,
> > nor anyone else can be expected to know his systems as well as he does
> > and it is inevitable that it will remain quicker and easier for him to
> > fix most issues - which in almost every case he does (I woke him at 3AM
> > yesterday for example). Similarly I wouldn't expect anyone but a member
> > of my staff to be able to fix an issue on a Vale server as quickly as I
> > might.
>
> As an appendum to this ... in over 10 years, what happened this summer is
> a first ... and only due to several things happening at once, the biggest
> problem not being that the server went down, but that I was in between a
> move ... had the server gone down either before, or after, the move, gborg
> would have been back up and running within 6-12 hours ... thanks to a
> friend, I was able to get the backup server onto a network, but it was a
> very slow network, one that it would have taken gborg >40 hours to upload
> from the backup server ...
>
> Right now, all postgresql.org related vServers are backed up to the local
> network, onto a second 64bit HP Proliant server, in case the one their are
> on blows up ... as well, they are all backed up to a backup server on my
> network here ... as well, over the next couple of days now that things
> have finally started to quiet down, they will also be backed up to a
> second *off site*, 64bit server, where they could come online very quickly
> in case all my servers happen to blow sky high ...

That's fine and dandy, but the fact remains: this would not have
happened at a professional hosting company. It wouldn't have mattered if
someone was in the middle of a move, because there'd be plenty of other
people who could take care of the issue. There would be backup plans in
place that involved adequate bandwidth. There would be well thought-out
diaster recovery plans. Etc, etc.

I'm not trying to slight you or the work you've done maintaining the
infrastructure around here, but the fact is that none of the above are
in place.

Now, we have two choices. We can try and obtain hosting from companies
that will handle everything for us so that if something fails, they're
on the hook to fix it and get all the services up and running. Or we can
make sure that we have enough servers hosted at different data centers
so that we can keep things redundant ourselves.

Well, I guess the third option is that we can continue along with a
number of single points of failure.

Even with the hosting that CMD and other companies are providing, have
we looked at what the disaster recovery plans are? If CMD burns to the
ground tonight, how do we get everything back online? Where are the
backups? Do we even know who to contact when <insert arbitrary
site/server> goes down? Is that documented anywhere?
--
Jim Nasby jimn(at)enterprisedb(dot)com
EnterpriseDB http://enterprisedb.com 512.569.9461 (cell)