Annnnd, done!

Hadoop has actually been up and usable for a while, but we spent some extra time moving and optimizing the MySQL instance that Hive and Oozie use.

The cluster is currently catching up on jobs, so expect it to busy for a while.

Thanks all!

On Tue, Feb 23, 2016 at 9:40 AM, Andrew Otto <otto@wikimedia.org> wrote:
FYI, this will begin shortly.


On Thu, Feb 18, 2016 at 2:08 PM, Andrew Otto <otto@wikimedia.org> wrote:
Hiya,

We’re ready to upgrade the Analytics Cluster to CDH 5.5.  To do so, we need to schedule a maintenance period during which we can stop all Hadoop related services.  This includes Hive, Oozie, Spark, etc.

I’d like to plan this for Tuesday February 23rd starting at 14:00 UTC (09:00 US east coast, 06:00 US west coast).  We’ve practiced this upgrade a few times in labs now, and I don’t foresee any issues.  I predict that it will take us no more than 2 hours to finish, but just in case I’d like to reserve 8 hours for this.

Please plan on not using the Analytics Cluster between 14:00 and 22:00 on February 23rd.  I will update this thread again when we are about to start, and when we are finished.

Progress is being tracked here: https://phabricator.wikimedia.org/T119646

What we get:


Thanks all!
-Andrew + Analytics team