[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Question about shared maildir storage


  This is what I already have:

  One of my customers has the following setup for their mail cluster:

Two dual processors Xeon 3.2 ghz machines, 2 gB ecc ram, RedHat Linux Advanced Server 3.0, Veritas Storage Foundation Cluster File System, netqmail 1.05+tls+smtp-auth, courier-imap, maildrop is the selected MDA, shared storage (3 TB) via gigabit fabric (HP MSA1000, ultra320 scsi disks 144 gB), no local users.

Both nodes can write concurrently in the shared storage, veritas storage foundation allows that. smtp/pop/imap traffic is layer 3 balanced and the two boxes can write concurrrently in the maildirs with no problem. Everything works as expected, so we have a high availability active-active cluster mail ((es)smtp/pop(s)/imap(s)) system.

  The problem:

They have almost 180.000 users. They use heavily imap, so they have a ton of mail stored. They also have dedicated machines for mailing lists managers (sympa & mailman). There are some mailing lists with a lot of suscribers (90.000+), so when anybody posts to those mailing lists, the *same* message gets delivered 90.000 times in our mail cluster. That is a lot of (wasted) space

  What I want to acomplish:

I want to lower the space requeriments caused by the mailing lists in our maildirs storage.. But I do not have a clear idea of how to accomplish this. These are some random thoughts:

* Use a MDA with a SQL backend so when the *same* message is delivered to thousands of maildirs only one is stored and symbolink links are created. When a user deletes this message with his/her MUA, delete the symlink. Courier-imap should also be hacked for this.

* Use a MDA with gzip-enabled maildir message storage. I do not like this approach.

If this is the wrong mailing list for this question, please tell me the correct one

  Can qmail-sql help me?

   Thank you for your support