MediaTemple – Moving Clusters

We have been using MediaTemple for a while to host Marketo’s public website. They have a great design sense and cost twice as much as Bluehost, but we hoped that the added expense would mean better quality.  Recently, we moved from their DV to GS services and got put on Cluster 2.  From that moment, all hell started breaking loose.  The site was sluggish and sometimes flat out down.  It was a disaster.  (A cluster-F#%k, if you will)

So I asked to be moved to a different cluster.  Unfortunately, this didn’t work at all.  MT cannot move you from one cluster to another.  I thought this was really bad architecture.  You have to be able to migrate people easily as a host.  If the technology doesn’t support it, then they need to seriously examine their architecture.

I decided to take the issue into my own hands and make the migration happen manually.  We didn’t have any wordpress, drupal or other service running, so it was just a matter of downloading the static site.  I created a new GS service so I could upload the files.  This created another architectural problem.  You can’t have two clusters serving the same site at the same time.  It’s one or the other. 

BAM!  That meant I would have some downtime.  DNS propogation takes 4+ hours typically and sometimes can take longer.  If I couldn’t have both sites working at the same time, then the old one would be down for 4+ hours during propogation.  That sucked, but I didn’t think staying on Cluster 2 was an option, so I said, “Well, I will stay up late tonight and mitigate the downtime!”

The MediaTemple interface is elegant and very understandable, but often takes too much time to update certain information.  I believe this is based on an architectural decision to make updates in batch chron jobs versus inline.  I know this is a highly complex environment and there are alot of moving parts, but this decision makes the experience really weird.  You make a move and then go eat a sandwich, then make another move, then each desert, then make another…you get the point.

In the last 48 hours, I have spent quite a bit of time with the technical support people at MT to get through this.  Although they were pretty nice and friendly, I definetely felt like I knew more about DNS than they did.  That scared me half to death.  Any tech support for a hosting company should either know what a HOSTS files is or be able to escalate to a senior person who knows everything.  They never escalated a single one of my issues, even though I was clearly stuck with several hours of downtime.  Why didn’t they escalate?  Was there an escalation path?  What is going on over there?

In the end, I am disappointed with how awful the experience was, but I got through it and now we are on the new system.  Cluster 4 is all I had hoped (so far!).  I just hope it stays that way.  I don’t plan on leaving MT, but I hope they invest more time in infrastructure/architecture.

It brings up a good UX truth:  No matter how awful the process, at the end of the tunnel people are happy with success.

%d bloggers like this: