New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
"Your load average is more than 2"
This must surely be a rookie question so I'm sorry for that.
I had a warning from my VPS provider that my load average is over 2. What's wrong with this? My CPU usage is averaging 50% so it's probably storage swapping that's causing it.
I guess some servers I have to limit resources myself in some way. How do I do that? How do I know what the limits I have are to work with are?
Perhaps I misinterpreted the message and they were just giving me a friendly heads-up. I thought I'd try to figure it out myself or ask here than make more of an ass of myself.
It's a VZ-SSD... I don't have much more info than this AFAIK. Maybe the vSwap is shared and that's causing the problem...
Comments
If you are using too much CPU time, use 'renice' to lower the priority of your processes. Generally, anything over '1' is taken as using too many resources by shared hosts. I've never owned anything at HostHatch, so I can't tell you.
Maybe run a tcpdump when it is happening and compare it to when the high load isn't happening? Most load problems I've seen are related to L7/application attacks (whether it be towards a web server or a VoIP server.) Throughput can be small and CPU usage won't look too high.
How many cores do you have?
As to swapping, how much ram do you have, and how much are you using?
What are you using the VPS for?
They might be friendly now but if it continues they likely won't be.
Providers usually don't talk without a reason and in your case this isn't the provider being chatty, it's a warning before suspension because you are using too much CPU. You need to stop asap if you want to continue using their service.
If you're using 50% cpu 24 hours a day, yes you will be suspended. You are sharing that hardware with dozens or hundreds of other people. It's fine to burst to 100% for occasional chunks of work and to use cpu when serving traffic requests, but most VPSs are near-idle most of the time.
If you want heavy cpu usage steadily, you need to get an unlimited-cpu VPS (they exist but cost more than typical budget VPS) or get a dedicated server.
I like pie. @willie what is your favorite pie?
If you paid for 7% of the pie, would you eat as much as you could for as long as you could before you were scolded, or would you eat the meager slice you paid for?
Oooor
No thanks. I don't want toppings.
What if you paid for "fair use" of the pie?
Ah I see. It's only one CPU, that's why.
Keeping it to 1 or less should be my goal I guess?
So, a VM system that balances the load fairly automatically is not possible? I guess as a result of that knowing how much resources are available for us to use is not possible because, I guess, people would try to chart when time is available and use the processing at that point, or game the system.
I will re-nice but that will only be within my VM won't it? I still have to limit processes to avoid load >1 right? I guess ulimit isn't enough for that? Is there something out there to help me keep load below 1?
All this was caused by one hungry process so I'm hoping this means it will be a simple re-nice.
One question is, what are you doing that's eating up CPU constantly?
Probably cpulimit
Renicing won't help at all with the load. It will be just more friendly with your other processes.
Long term load of 1 means 100% which is bad. Short term load of 1 is usually no problem. What are you doing that uses so much cpu? If you actually need it, then I can tell you that having a dedicated server is a liberating experience, since it means you can use all the cpu you want.
AR lets me bust a 20 load for days before they say anything. It's strange.
Fair use doesn't usually mean "lol nobody's looking its all mine", unless you are an incredible dickhole.
It can be ok depending on how the provider sets things up. BuyVM (KVM slices), Wishosting, and the "cloud" providers specifically allow it. Your cpu share does get cranked down over time if you hog it, but whatever you can use is yours.
Did I mention what a joy it is to have a dedi if you're a CPU hog? I've run multi-week jobs at 100% of all 4 cores on my $30/month Hetzner i7 auction box. That's as much compute power as around $500/month worth of EC2 instances last time I looked. I highly recommend it and don't know why so many people pay those ridiculous AWS bills.
We also allow full utilization of resources
try ramnode vds if you would like to cpu abuse full time.
Thanks for your help guys. I'll switch to a cloud provider as I like that adaptation to use.
I was just trying to run a rippleD server as a test but it seems like a total resource hog so I'm not sure if I want to do it. However, I'd like to be able to play with things felt time to time so it's good to know what's available to use
Just get any KVM VPS, you won't have the load issues like on an OpenVZ. Regardless you should look into what's causing it... no point of ignoring it.
Daemons for virtual currency, such as Ripple and Bitcoin, are forbidden by most VPS providers. Sometimes 'Cloud' providers can also suspend your server.
It is recommended to use a dedicated server. If you need to use VPS, I recommend KVM or XEN based VPS instead of OpenVZ or LXC.
Hourly billings, elastic creation and destory, places other than Germany, AWS IAAS intranet connections, etc.. For example, there's no choice so far for a budget high performance dedi in US. The closest thing is OVH BHS.
But I'm sure lots of people are using AWS the wrong way.
Hourly billings except that AWS costs as much in 2 days as the Hetzner server does for a whole month. There's reasonable comparable dedis at places like Nocix. There are various services that AWS has that you can't do on a budget dedi, but pure computation isn't one of them. I've run up some pretty big AWS bills at work doing pure computing (data analysis) that could have been done on a Hetzner, but we spent the money at AWS because we were an AWS shop.