We implemented the clustered configuration of the community software today, and after two hours it continues to run smoothly. We will continue to monitor it closely. The clustered configuration will allow the site to handle peak traffic (early morning weekdays, Pacific time) better, increase cache sizes, and restart a single node without affecting overall site availability.
One note: As part of the cluster upgrade, we increased the expiration time for the user status level cache from 15 minutes to 1 hour. This means it will take up to an hour for status levels (i.e. points) to be recalculated and displayed. We made the change because we are still researching a bug in the status level calculator that has caused threads to hang -- by exercising this code less frequently, system stability increases. We are of course working on fixing the root cause, which will allow us to decrease the expiration time again.
As for the user interface fixes with the 2-column layout, we decided to wait a few more days before releasing them in order to minimize risk in rolling out the clustered configuration.
JohnTroyer
ericni
RDellimmagine
Badsah
