+-Le 29/12/2005 14:52 -0500, Dave McMurtrie a dit :
| Nikola Milutinovic wrote:
|
|>
|> Are we talking about cyrus murder? I'll definitely take a look.
|
| He's actually talking about the replication code that David Carter wrote.
| Murder only offers scalability, and does not give you any type
Igor Brezac wrote:
On Wed, 28 Dec 2005, Nikola Milutinovic wrote:
Hello all.
How would one go about building a Cyrus cluster?
Suppose you wanted to make two servers a failover IMAP cluster. The
way I imagine it, the MTA can forward messages to two servers.
Problem is, deleting from one se
Nikola Milutinovic wrote:
Are we talking about cyrus murder? I'll definitely take a look.
He's actually talking about the replication code that David Carter
wrote. Murder only offers scalability, and does not give you any type
of replication of the backend mailstore.
Thanks,
Dave
C
Nikola Milutinovic schrieb:
> How would one go about building a Cyrus cluster?
>
> Suppose you wanted to make two servers a failover IMAP cluster. The way
> I imagine it, the MTA can forward messages to two servers. Problem is,
> deleting from one server will not reflect on the other. I'd like to
On Wed, 28 Dec 2005, Nikola Milutinovic wrote:
Hello all.
How would one go about building a Cyrus cluster?
Suppose you wanted to make two servers a failover IMAP cluster. The way I
imagine it, the MTA can forward messages to two servers. Problem is, deleting
from one server will not reflect
I don't know if in your case a cluster is the better solution. If i
should evolve my cyrus configuration i would evaluate aggregator
http://asg.web.cmu.edu/cyrus/ag.html as first possibility. It's a
solution conceptually in the middle between perdition and a cluster.
One point of strenght of ag
We use drbd and heartbeat on the backend mail servers (active/passive,
data is real time replicated from active->passive). Has worked very
reliably for several years, however it is not the most clean solution.
I've heard / read bad things about GFS based shared storage (cyrus wiki
actually has