|
/
Zope
/
Apsis
/
Pound Mailing List
/
Archive
/
2004
/
2004-05
/
hot failover for pound itself
[
Content-Aware or Non-Content-Aware? / ... ]
[
Memory Leaks / "Alexander Meis" ... ]
hot failover for pound itself
Paul Chvostek <paul+pound(at)it.ca> |
2004-05-12 21:17:36 |
[ FULL ]
|
I'm looking to replace a pair of Cisco CSS 11150 units that are no
longer supported by Cisco. Rather than pay thousands (and thousands)
to Cisco for a more recent CSS, I'd love to implement a pound-based
solution.
The thing I need to cover that isn't already part of pound is failover.
This obviously isn't something that can easily be added to pound itself
because it's largely operating system dependent, and I'm curious how
others have tackled this.
My operating system of choice is FreeBSD. Has someone implemented pound
in conjunction with freevrrpd? I can consider other OSs as well, but
obviously nothing bleeding-edge. Are there solutions available that
don't include a single-point-of-failure?
Thanks.
[...]
|
|
|
Re: hot failover for pound itself
Chris Gamache <cgg007(at)yahoo.com> |
2004-05-12 22:18:27 |
[ FULL ]
|
Hey Paul, here's my $0.02.
You're right in that Pound will never support HA within itself. It is only a
Load-Balancing-Reverse-HTTP-Proxy, and will continue to be only that! :)
We use Heartbeat (http://linux-ha.org/) to
maintain HA between pound servers.
(There was talk of a port to FreeBSD at one time or another... :) When the
primary Pound goes down, the secondary Pound steps up to take its place. When
the primary Pound comes back up, the secondary Pound gets pushed back, waiting
to take over once again. We use daemontools (http://cr.yp.to) to manage Pound.
That makes it a snap for Heartbeat to bring up and take down Pound. I have to
admit that the only time the secondary Pound has been used is when we've pulled
the plug on ourselves for testing. We've never had a problem where the Pound
server went down on its own. You really ought to get two (or n) boxes set it up
and make sure that it meets your needs for HA. Unlike the HA/LB black boxes out
there, it won't set you back $$$$$ for a test-run or two.
CG
--- Paul Chvostek <paul+pound(at)it.ca> wrote:[...]
[...]
|
|
|
Re: hot failover for pound itself
Robert Segall <roseg(at)apsis.ch> |
2004-05-13 11:00:34 |
[ FULL ]
|
On Wednesday 12 May 2004 21.17, Paul Chvostek wrote:[...]
We've done this a few times with simple shell scripts and a bit of arp
spoofing. OpenBSD, stock Pound. I believe the latest OpenBSD (3.5) has even
better facilities for this, but we haven't tested those yet.[...]
|
|
|
|