Forum: Mongrel Mongrel Cluster: Rolling restart?

Hi all,

Forgive me if this has been asked before, but is there a reason
mongrel_rails cluster::restart stops all ports and then restarts them
again?  Wouldn't it be better to restart each port one-by-one in
series so that upstream proxy servers can fail over while the restart
is occurring, or am I missing something?

Thanks,
eden

On 11/3/07, Eden Li <eden@hulu.com> wrote:
> eden
Eden,

You are not missing anything, this is the case. I use my own custom 
cluster
script which does exactly this instead of mongrel_cluster.

That said, patches are welcome :)

Thank you,

  ~Wayne

> Forgive me if this has been asked before, but is there a reason
> mongrel_rails cluster::restart stops all ports and then restarts them
> again?  Wouldn't it be better to restart each port one-by-one in
> series so that upstream proxy servers can fail over while the restart
> is occurring, or am I missing something?

You might also like Seesaw if you don't need to run any complex
migrations:

http://synaphy.com.au/2007/8/20/seesaw

Regards,
Andy Stewart

mongrel_rails and mongrel::cluster are slated to all change in 1.2.

Evan

Here's a patch to Cluster::Restart based on some code that Eric Kolve
sent to me (I added check_wait):

http://rubyforge.org/tracker/index.php?func=detail&aid=15427&group_id=1306&atid=5147