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.
Is overselling affects vps load?
Hello to all
i have a vps from a hosting company
i start to using the openstatus monitor script
i set up the server on a third vps and 2 clients in my 2 other vps
i got about 30! email alerts per day from the small one vps i monitor that the load is to high
with values 1.5, 1.2, 1.7 even with 2.+
I want to ask from your experience
The high load menas that the vps hosting company oversell to much ?
The high loaded vps has nothing on it except some rsync scripts that runs once a day
and the memory usage is 19MB of 256Burst
Virtualization is OpenVZ
Thanks
Comments
Well, what constitutes a high load average depends on your VPS. A load average of 2 on a VPS with 1 core is high, but a load average of 2 on a VPS with 4 cores is actually low - it can handle twice the load, in fact. So, the first step would be to change your warning and critical levels in your openstatus-client.conf file to reflect this. I typically use warning = 1.5 and critical = 2 on a VPS with 1 core.
The next thing to consider is if this is sustained load, or spikes. Load average spiking is pretty common and not really cause for alarm, but a high load average sustained over 5 minutes or more probably is.
The vps has 1 core.
30 spikes per day?
In as short way as possible, on OVZ it can happen.
It's not uncommon for there to be that many spikes. Consider a process that spawns 3 or 4 subprocesses during its execution, and each those subprocesses take a few seconds to complete. All of a sudden, you have 3 processes waiting in the run queue, and you're going to see a load average spike.
With 1 CPU, a load of 1.5 is high but not too high.
Even up to 2 load, it should still work ok for your rsync scripts.
If all you are using it for is rsync and you are getting a good price, I wouldn't worry about it.
Change your monitor script to report a load of 2+ and enjoy more of your day than worrying about CPU load on an rsync server.
We have had a node with a single client hit 80.0 load so high loads indicate usage, not overselling. Even undersold servers will hit 5000% if a single VPS has an out of control process.
We allow our VPS nodes to burst just over 10.00 for a few minutes but anything over that amount constantly Is not good at all!
Majority of our servers can handle much much more stress but we do not push It to threshold, All our servers are the E3-1230 sandybridges (8 Log Cores) @ 3.2Ghz so they can handle good.
But, a really overloaded VPS node will make a basically empty VPS have higher load avgs than the same VPS on a empty node.
I have some Platinum boxes running at 178 load for 8 hours on end and the rest of the node is unaffected!
What am i saying is that my vps does not use the load.
I even removed tiger and rkhunter and stil got alerts for high load.
So it is possible...
@JustinBoshoff do you use openvz?
only to share my story. A company Im working for, there were a few virtual servers on the node. One of them hgad load 4 and mysql process was taking 200% CPU. It caused another VPS load to go over 40. When I stopped the first VPS(loaded to 4 - 200% CPU mysql process) everything came back to normal. Real world experience, I was never able to explain that
@apollo15: which virtualization?
@bobinfo
Its OpenVZ.
@apollo15
Yes Openvz, If you configure your containers to only use a percentage of the core/s assigned and not the hole core/s then it wont impact other containers as much because there will be some processing power left so the other containers don't need to wait to be able to execute.
The reason you would get the 40 load on the non busy container is because container1 can use 100% of the cpu so container2 has to wait for container1 to finish before being able to execute. If you configure the containers to only have 80% of cpu container1can run at 100% of allocated cpu and container2 can still execute.
I got 1 core(100%) per VPS, so basically more VPS use the same CPU, thus if one VPS pushes the CPU another one sees that. Right?
Thanks for explanation.
That's correct.
Ok, if you are running rsync it is probably going to have a load average spike as soon as that process kicks up because you are putting a ton of IO load on the server all at once.
How do you get an indication of the load that other OpenVZ VPS users are causing?