Cyrus 2.3 on shared filesystems

[Ken Murchison]

Thomas Vogt wrote:

> Am Mittwoch, den 03.11.2004, 16:20 -0500 schrieb Ken Murchison:
> 
>>>IMAP2, so IMAP3 can't access mailboxes on STOR1. Does this work?
>>
>>No.  You're talking about two levels of synchronization here, the 
>>current Murder code only handles the following scenarios.
>>
>>
>>Standard/Unified Murder:
>>
>>+-----+ +-----+ +-----+ +-----+
>>|IMAP1| |IMAP2| |IMAP3| |IMAP4|
>>+--+--+ +--+--+ +--+--+ +--+--+
>>    |       |       |       |
>>+--+--+ +--+--+ +--+--+ +--+--+
>>|STOR1| |STOR2| |STOR3| |STOR4|
>>+-----+ +-----+ +-----+ +-----+
>>
>>In a standard config, the mailboxes.db on each IMAP server only contains 
>>those mailboxes contained on its local STORage.  The master mailboxes.db 
>>(which contains ALL mailboxes and the servers on which they reside) is 
>>located on the MUPDATE master and slaves (frontend).
>>
>>In a unified config, the master mailboxes.db is located on the MUPDATE 
>>master.  Each IMAP server is an MUPDATE slave (no frontends necessary) 
>>and differentiates between local and remote mailboxes.
> 
> 
> Interessting.
> 
> So if I've something like this 
> 
>           {mail.domain.com}
> +-----+  +-----+  +-----+  +-----+
> |IMAP1|  |IMAP2|  |IMAP3|  |IMAP4|
> +--+--+  +--+--+  +--+--+  +--+--+
>    |        |        |        |
> +--+--+  +--+--+  +--+--+  +--+--+
> |STOR1|  |STOR2|  |STOR3|  |STOR4|
> +-----+  +-----+  +-----+  +-----+
> 
> Lets say, I only host 1 Domain on all this Servers. Stor1 has user from
> a to f, Stor2 from g-l etc. For outside Users everything transparant
> combined as mail.domain.com. So Users don't see all this different
> stores. Will this work? 

Yes.


> I mean if its transparent user A will perhaps connect with
> mail.domain.com to Imap3. But his mailbox folder is on Stor1.
> 
> Can someone explain how this will work?

The whole point of the is to split the mailboxes on to a bunch of small 
(cheap) servers but still present a unified namespace to the client 
(unlike a Perdition cluster).

The big difference between the standard and replicated config, is that 
in a standard config you have dedicated frontend machines (proxies) 
which direct the client to the correct backend(s).  In this config, the 
backends never really talk to each other.

In the replicated config, the proxy functionality has been merged into 
the server code, so you don't need dedicated frontends, although you can 
still use them if you like.

-- 
Kenneth Murchison     Oceana Matrix Ltd.
Software Engineer     21 Princeton Place
716-662-8973 x26      Orchard Park, NY 14127
--PGP Public Key--    http://www.oceana.com/~ken/ksm.pgp
---
Cyrus Home Page: http://asg.web.cmu.edu/cyrus
Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html