On 5/10/06, Brian Kotek <[EMAIL PROTECTED]> wrote:
Chris, I was just asking about this on the CFCDev list. I'm just not sure
how to handle a situation where I don't have direct control over the servers
and an upgrade breaks something that requires the template cache to be
purged. Do we have any idea why this happens or if there is any way to deal
with it (aside from manually purging the cache in the CF admin)? Thanks.

Not sure what the issue is here. adobe.com very specifically relies on
the template cache in order to be able to do rolling upgrades with no
downtime. We run a build (code update) to the cluster and "drain"
sessions on half the cluster, then cycle those servers and do final QA
and put them back into the cluster. Then we "drain" the other half,
cycle those and put them back.

The point is that the template cache allows us specific control over
when the server picks up the changes.

On 5/10/06, Chris Scott <[EMAIL PROTECTED]> wrote:
> Ahh, template cache. Always be VERY careful with your cached templates on
production servers (and clear them!). Cached templates will cause massive
headaches when working with all the major frameworks when you do any type of
upgrade/deployment ...

adobe.com uses Mach II heavily and this point goes beyond the trusted
cache - you need to refresh your applications as well as clearing the
cache, because you have "old" CFCs in memory even if you flush the
trusted cache. That's an application specific piece of code you need
in order to be able to flush the *application's* 'cache' of CFCs and
recreate them (Mach II, Model-Glue and Fusebox all have hooks to do
this for their own code but application authors need to bear this in
mind too!).
--
Sean A Corfield -- http://corfield.org/
Got frameworks?

"If you're not annoying somebody, you're not really alive."
-- Margaret Atwood

Reply via email to