[Gluster-users] Replication logic

Gionatan Danti g.danti at assyoma.it
Sun Dec 27 21:49:29 UTC 2020

Il 2020-12-27 21:58 Zenon Panoussis ha scritto:
>> For such a project, I would simply configure the SMTP server do to
>> protocol-specific replication and use a low-TTL DNS name to publish
>> the IMAP/Web frontends.
> Either you know something about mail servers that I would love
> to know myself, or else this idea won't work. That's because
> even if you could configure three SMTP servers as relays to
> each-other (which you cannot), IMAP actions (move, delete etc)
> do not propagate. So, one day you organise your mail nicely
> in folders, next day you find all of it back in your inbox.
> Or you go looking in your Sent folder for that mail you sent
> last week, and it's just not there. That kind of thing would
> probably frustrate everyone much more than a bit of latency.

Uhm yes, what I wrote was quite confusing. I was really referring to 
something as that: https://wiki.dovecot.org/Replication

Re-reading your original question, I think geo-replication should work: 
after all, if I understand it right, you would only use the 
remote/backup copy in case the primary/local server goes down. Right?
However, you should carefully test if Gluster is, by itself, capable of 
the load you are going to put over it: considering that a mailserver can 
easily store millions of small files and Gluster relatively low file 
lookup performance, I would thoroughly test it before putting this setup 
into production.

About the closing consideration about a "bit of latency", I think you 
are *way* over-optimistic for replica volumes: Gluster sync replication 
is very latency boud and, if things did not changed recently, it was 
strongly advised against running sync replication between WAN links.


Danti Gionatan
Supporto Tecnico
Assyoma S.r.l. - www.assyoma.it
email: g.danti at assyoma.it - info at assyoma.it
GPG public key ID: FF5F32A8

More information about the Gluster-users mailing list