librelist archives

« back to archive

Lamson redeployment

Lamson redeployment

From:
Yang Zhang
Date:
2010-07-21 @ 18:42
Any advice/experience on redeployment with Lamson? RELOAD simply
reloads the top-level handler module, and not any imported modules
(unlike e.g. Paste's --reload). Any simple way to reliably queue
messages while stopping and starting the handler? Is the recommended
approach to use a separate mail server (perhaps postfix) in front of
Lamson, with a fast retry? Thanks in advance.
--
Yang Zhang
http://yz.mit.edu/

Re: [lamson] Lamson redeployment

From:
Zed A. Shaw
Date:
2010-07-21 @ 19:11
On Wed, Jul 21, 2010 at 11:42:58AM -0700, Yang Zhang wrote:
> Any advice/experience on redeployment with Lamson? RELOAD simply
> reloads the top-level handler module, and not any imported modules
> (unlike e.g. Paste's --reload). Any simple way to reliably queue
> messages while stopping and starting the handler? Is the recommended
> approach to use a separate mail server (perhaps postfix) in front of
> Lamson, with a fast retry? Thanks in advance.

That's what I do.  Postfix is really easy to setup if all you want it to
do is receive everything and put it in a maildir.  Then you have lamson
pull from the maildir using a QueueReceiver.  That lets you do full
restarts whenever you want.

Another point though, is I tend to not do live restarting/hacking on a
running server unless its an emergency.  I do my work on a local
machine, have unit tests and things to make sure it works, then deploy
at a period of down time and very carefully.  No matter how a system is
architected a restart always has a probability of dropping a message, so
you want to do what you can to reduce that.

-- 
Zed A. Shaw
http://zedshaw.com/