So, my excitement about the load averages was short lived, as they’re inaccurate. Not that big of a deal, but dealing with CloudLinux definitely takes a bit of an adjustment since the default indicators I’m used to relying on are simply not the same.
I’ve had it installed for nine days, and the only issue I ran into is that I was quite surprised to find myself dropped in my own container. Maybe it’s ego, but it’s my server and if I want to overload it I should be able to overload it, darn it. OK, maybe not. Maybe this is a good thing. [grumble grumble]
For the most part, this isn’t a problem, and I’ve run into nothing I cannot do, but I haven’t had to think much about who “owns” the process when I’m doing something, and as I consolidated my first two servers, I was shocked to find myself locked out of ssh during the move because the move took place through my user id, and unpacking and moving obscenely large sites from one server to another is sickeningly resource intensive. I kept my CLI until I dropped it for some reason or another, and tried to log back in, and then it was a no go until the site I was moving finished moving.
I’m still having some issues wrapping my mind around it all, but it’s slowly coming.
The consolidation was done two days ago, and the TTL had been taken down so that the new sites were resolving within about 20 minutes to their new server. Rather than using load, I’ve been relying on iostat to keep track of the situation and so far, I’m fairly impressed.
This is only a dual dual core (it’s been around a while), and the servers I’m looking at replacing in the 2-1 consolidation are a fair bit more robust than this one. I likely could get a Westmere with an outrageous amount of RAM and really pack ‘em on, but disks blow up and the restore on a server with Terabytes of data is enough to give me the willies, so regardless of whether I could go higher than 2-1, I am sticking with 2-1. (Heck, my 2-1 doesn’t get close to some companies server population ceilings that I could name. I could name them. But I won’t.)
For those curious, our account ceiling per server was 350, and now it’s about 650-700.
We lost the Daily Process Report in cPanel, which annoyed me a bit but then I realized that the usefulness of that, since everyone’s in an LVE, is somewhat questionable. Just decided it didn’t really matter all that much. If people’s crap coding and overloading is contained, they can have at it now. Bang away. No more chasing after people with fines and lectures. They’ll know exactly when they’ve pushed too far – their site will stop serving. Yay for making my life easier.
It appears that the bandwidth use was totally reset in the reports for the month when we installed CloudLinux, so my clients get a gimme if they went over. It does appear to be accurate with regards to going forward, so that’s cool. If you know how to recover that, don’t bother to tell me. I don’t much care. 
But this server, because it had huge RAM numbers, had a few of our highest traffic clients (keeping in mind we have awesome clients, and our high bandwidth customers tend to be 75 Gig/m WordPress Blogs). I’ve checked the response time of those higher resource/higher bandwidth clients multiple times and have seen speedy responsiveness at all times of the day and night, all resource use levels. Apache error logs don’t seem to be erroring on anything abnormal.
So far, I am pretty durn impressed.
