High Server Load

[qsgt]

Hi,

 

I've been around HelioHost for a few years now under a few different forum user names. I've used HelioHost for some personal stuff and to host an event registration service for a charity that I work for. We are about to publish a registration form for an upcoming event and I'm slightly nervous about doing it because Tommy's server load is the highest I've ever seen it and I'd father not have a massive crash while we accept event registrations. Tommy just went down for half an hour as I was getting ready to open registrations. See below:

 

 

 

Believe me, I know that this service is free and I'm getting what I pay for. Your service is absolutely fantastic!! It lets me learn about web design while saving my charity money (We would pay for hosting of the registration system if we could).

 

This post is in no way meant to be a criticism, I'm just curious to see if you know why the load was so high and what the likelihood of it being a regular occurrence is?

 

I've also seen this question asked many times before but what does the load figure actually represent? I thought accounts started getting suspended when the overall server load exceeded 20?

 

Thanks for your time, Keep on being awesome!

qsgt        

[wolstech]

That was abusive use, and two people did get suspended for it. The problem is that when the load is massively high like that, it can take a significant amount of time for the suspension script to actually work. On a normal day, suspending an account takes 30s-1m. When buried in load, that process can take an hour or more. Then there's the fact it's not always just one user needing suspension.

 

Most of the downtime period is the time spent actually doing the work to suspend the user. Once that process reaches the point where it kills all the user's processes, that's when the account no longer produces any load, and the server load starts to return to normal.

 

The give away of a severe high load incident on that chart is the almost instant cutoff from green to red where it goes down, followed by the load tapering from red back to green over a few minutes at the end of the event (as the abusive processes die).