[Labs-l] Additional labs capacity and rebalancing

Andrew Bogott abogott at wikimedia.org
Mon Jan 5 06:21:25 UTC 2015


     We added a couple of new extra-large nodes to the eqiad virt 
cluster at the end of last month.  They've been performing fine, so this 
week I'm going to do a bit of re-balancing.  I will be moving a few 
instances off of overloaded servers (virt1002, virt1005, virt1009) and 
on to the new, empty servers.

     In most cases you won't notice these moves -- migration will cause 
a few-minute freeze on the instance but will not require a reboot.  On 
occasion an instance doesn't survive the move intact and needs a reboot 
afterwards.  If you have an instance running on one of the above hosts 
that cannot survive a 10-minute interruption, please let me know; 
otherwise, no action is required on your part. I don't plan on moving 
any tools nodes unless absolutely necessary.

     Note, also, that the addition of these servers means that our space 
crunch is largely at an end.  If you need a quota raised and we've been 
putting you off, now is the time to re-ask on IRC :)

-Andrew




More information about the Labs-l mailing list