Well, Hugo, don't know if the list server is messed up, and it reposted this
for you, or if your system reposted it, but, on the chance that you didn't
see my answer, I'm reposting it.
mark
On Thursday 10 October 2002 11:02 am, you is done writ:
> When we implement a failover recovery using the heartbeat protocol, it
> really seems to work and for that I haven't any doubt (well I haven't tried
> yet, but...), but I think there is a problem in keeping the DBs
> synchronized.
> Having two servers (master and slave) running slapd and slurpd and the
> heartbeat, and assuming that all the physical redundancy is maded:
> when the masters goes down the heartbeat puts the slave being the master for
> providing updates to the clients (everything is ok! till here), but, when
> the master comes up the heartbeat puts everything like the original
> configuration and what about the updates maded to the DB?
First, let the secondary (slave) act as the primary (master) once you've
failed over. They *should* be identical, anyway. Also, a heartbeat is really
nothing more than the secondary pinging the asserted virtual IP of the
primary. It, in itself, does no updating.
Second, you should either fail *back* over to the primary (say, by pushing
the reset switch), or by turning off the secondary, and then bringing the
service on the primary back up, then allowing the secondary to look at the
heartbeat again.
All of this should be done by the high-availability "service". If this is
only a home-made (please note, not "maded", made is the past tense of make)
system, what you need to do is have a script that looks at the heartbeat (the
once-every 1/5/10/whatever second ping of the asserted IP (both servers
should have their own, default IPs, *and* an asserted virtual IP), running on
both machines. If one, coming up, sees that IP being asserted by the other,
it should sit there and loop, watching for it *not* to be asserted. That, of
course, would happen if the other machine goes down. Then the script should
bring up you LDAP server and replicator.
That's how we did it, using DEC, er, Compaq, uh, HP's commercial
high-availabilty hardware and software for the City of Chicago 911 system.
mark roth, currently job hunting in Chicago (hint, hint)
<[EMAIL PROTECTED]>
--
Certified lunatics are shut up because of their
proneness to violence when their pretensions are
questioned; the uncertified variety are given the
control of powerful armies, and can inflict death
and disaster upon all sane men within their reach.
--Bertrand Russell, Power [1938] ch.16
--
redhat-list mailing list
unsubscribe mailto:[EMAIL PROTECTED]?subject=unsubscribe
https://listman.redhat.com/mailman/listinfo/redhat-list