I have the mirror system set up as three tiers:
Tier I. Primary activities server.
Tier II. Primary download node.
Tier III. Secondary download nodes.
Aside - we refer to addons as activities
The Tier I server is set with MIRROR_DELAY = 30. All new activity are
served from this server for the first 30 minutes. The Tier II server
(with current traffic loads, we only need one) is set with a priority
1. (very low) This server handles new activities older than the 30
minute delay but before they are mirrored elsewhere. The low priority
means that if an activity is available elsewhere it will be grabbed
form there. Tiers III is the general availability mirrors.
It looks like we will be able to tune and scale up the download
infrastructure by adding machines to one of the three tiers and tuning
the MIRROR_DELAY and sync frequency of specific mirrors.
Finally, my next set of question:)
Is there a specific memcache hit rate target that you aim for? Is is
possible to use multiple memcache machines? In our case, we have
limited memory on our production machines but a lot of memory on a
machine which is used as a vm host for developer machines. All
machines are on the same rack in our data center. Is it possible to
host some of the memcache on the production machine and create a vm to
specifically host additional memcache? How much memcache do you use
for the primary AMO sight?
thanks
david
We don't have a target, we aim for as high as we can get. :)
$memcache_config is an array, you can put as many machines in it as
you want. Use the 'weight' element to give one machine more traffic
than another.
Our AMO cluster has 3 memcache machines, each getting between 90-95%
hit rate. Most of the other settings are the defaults.
Wil