Re: Failover
On Mon, 21 Feb 2005, Tom Brown wrote:
> On Monday 21 February 2005 11:57, Curtis Vaughan wrote:
> > Has anyone set up failover scripts on a debian box?
> >
> > I'm looking at linux-ha.org and in particular
> > http://failover.othello.ch/ but there is no debian specific package.
> > Of course, there is a source package, but I would be surprised if no
> > one in the debian community has looked into this before.
> >
> > Let me explain what I want to do.
> > We want to have a failover mail server.
> > As I understand it should be fairly easy to do the following:
> > 1. Have the primary server rsync over any new info to the failover
> > server immediately.
if the primary server died ... how can you rsync anything ?? it's dead
and you're s.o.l.
for mail server failover ...
use MX records ... it does it all automatically
just be sure that each mail server is configured with
the same antispam/antivirus/other-ids stuff
if you want to copy/rsync mail1:/var/spool/mail/users onto
backup-server:/var/spool/mail, than that backup-server
should be updated each time an incoming email is received
which is trivial to do .. it's lot trickier to delete
the emails that have been read, since the pop daemon
may or may not be part of the mta ..
for mail pop server .. ( for people that used netscape/ie/mozilla/etc )
you can use 2 or more pop servers that looks like 1 server
but the bottom line is that each pop server will have to
sooner or later go to ONE place in the database of pending
emails to say that email#12002 has already been read/pop'd
( it still has a single point of failure -- the database )
if you use pop1 for engineering
and use pop2 for finanance
and use pop1000 for everybody else, than some folks will
be offline, when pop1 dies, but they can use pop2 to
get their emails
- there is NO single point of failure
ha and rsync and network backups is an overkill and creates lots
of extra work ...
- what is the loss of productivity if the mail server goes down ?
- what is the loss of productivity if the pop server goes down ?
- what is the loss of productivity when the "mail system" doesnt
work right for sending/receiving emails and pop servers and
duplicate messages and antispam/antivirus filters on each server
=== once the primary server dies .... you CANNOT rsync anything anymore ..
it's dead, but than again is it dead because the network cable got
pulled or is it dead because the disk died or dead because exim died
lots of fun stuff
c ya
alvin
> > 2. Have the failover server monitor the primary server. In the even
> > the primary is unavailable the failover takes the IP address of the
> > primary and becomes the primary. When the primary comes back on, it
> > detects that the failover is now the primary and the former primary
> > now becomes the failover.
> >
> > That's basically it.
> >
> > Curtis
>
> What you describe is what drbd and heartbeat do. I implemented both on a
> couple of Debian Woody servers and they work out great. I would
> recommend getting the latest source and not the debian packages.
>
> Hope this helps,
> Tom
>
>
> --
> To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
> with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
>
Reply to: