Hiya folks, I thought maybe somebody could shed some light on this for me. I used the migration function maybe a week or so into March, and everything went pretty smoothly. But it wasn't until today that I happened to notice that the CPU usage graphs had a sudden change ever since the migration. It used to float around 2%, but now it sits at around 5% and never goes below it. Disk IO also sits permanently higher than before.

My only assumption as for the CPU usage might be related to how Linode measures it, and perhaps there's more cores in the new host machine and that factors in? But I don't see how that affects the disk IO, though. Maybe it's just all measured differently on a Xen server?

Here's a snapshot of the graphs:



That spike in network traffic is just from where I did a backup before the migration.

So yeah, I'd appreciate any insight anyone can offer on this. Maybe it's nothing, but I'm curious enough to find out nonetheless!

Heh, I just openened a ticket about this yesterday because I noticed the same unusual (to me) CPU usage. This is what I was told:



I'm also guessing the difference in scale is due to a switch from UML to Xen, which is what my Linode went through.

Just for reference, here are my charts:



edit: Umm.. now I see that our Disk IOs do look very different though. Seems like we got migrated right about the same time. Were you in Dallas 5?

Looking at your graphs I wouldn't worry about it, mine look very similar to yours, (apart from once per day I have a huge spike in cpu/io/network as my node backs up).

Like the support ticket said you're graphs are using small increments on a big graph so it looks worse than it is

Disk I/O is indeed measured differently between UML and Xen.

I dunno about CPU, but it doesn't surprise me that usage is a little different after making such a big change.

_________________
Matt Nordhoff (aka Peng on IRC)