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.
What's the standard of abusing I/O?
AvelineSwan
Member
Comments
http://www.lowendtalk.com/discussion/3306/do-you-know-any-way-to-continuously-check-iops-suspended-for-abusing-io
Sounds interesting!
It depends on the hardware. A VPS doing 100 iops on a node with a single sata disk that's only capable of ~200iops is an issue unless there's only 2-3 VPSs on that disk, but a VPS doing 300 iops on a node with SAS RAID10 shouldn't be an issue
And I am still waiting for something to tell me how to do it manually, because I don't want to use that observium/munin/any-monitoring-system
There is something at /proc to watch?
Actually you have to use a tools to monitor it continuously to calculate the IO speed, so ...there's nothing at /proc to watch.
See here: http://wiki.openvz.org/Category:Monitoring
@aveline I don't see anything there saying that.
Perhaps have the provider run a benchmark when empty, so they can set the 'maximum possible IOPS', then, once the user is using more than x% of the maximum, suspend?
Apologies if I sound crazy.
I think notifying user and lower their I/O priority for a while will be greater.
Suspend could cause some conflicts.
I like how securedragon just shuts down your VPS and asks you to fix the excessive resource usage. It'd be better than just suspending the VPS.
@Corey I mean "There's no method to do it manually", and the page I offered is just used to prove it.
@aveline You can use atop to see how much % of disk IO a process is using. If the disk is pegging out on IO (which you can see in atop) then you can guess about how much IO that process is using if you look at how much IO is going on at the disk at that time. Then you can use the vzpid command to link the process to the container.
The page you offered does not prove that you can't do it manually (or at least I can't find it), because I ONLY do it manually.
@Corey Awesome. Thank for your suggestion.
@aveline I am actively looking for a better solution so if you find one let me know
There must be a method.
That bloated monitor software obviously reads something from the kernel and processes it in some way. So, I must be able to do the same in bash or whatever language I like.
Look at the source code man! :P