The mailing list for listmasters using Sympa

[Amos <address@concealed>]

What about a hybrid approach?  Have the spool state information in the db, but the actual message be stored on the filesystem?  Both CMU Cyrus and Zimbra do that.

On Tue, Apr 21, 2015 at 5:22 PM Steve Shipway <address@concealed> wrote:

Soji wrote:

> IMO using RDBMS for spool is not good idea.

 

I would agree with this, except for the fact that, if you want a really high throughput, spool in the DB allows you to split the distribution and reception threads over multiple servers, so you can expand horizontally.  We have not needed to do this (yet), but sites such as Riseup have done so as they have a massive throughput.

 

While we can happily move to 6.2 and filesystem-based spools, it may cause issues with very large scaling of Sympa?  It depends on how the distribution thread locks the spool messages during processing to prevent other distribution nodes on a shared filesystem from taking them, and if there are any race conditions.

 

Steve

 

Steve Shipway

address@concealed

(GNU Terry Pratchett)