All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Causing high load
Greetings,
I'm currently hosting a few sites on a reputable VPS provider (I pay roughly $25 /m for 1GB, 4 shared cores, OVZ) and it'll display strange CPU loads as if it's waiting on i/o.
http://i.imgur.com/5zEmldA.png
To be honest with you, I haven't had this issue with for example, RamNode (on a single shared core).
This VPS is currently running the same config I use on 2 other dedis, and some low end VPSs without issue. Apache2, mod_pagespeed, nginx as proxy, memcached, DSO or FCGI, eaccelerator or APC. CSS/JS/images are served via EdgeCast.
So what's going on here? Node seems way overloaded. Should I contact them?
To note: this happens regardless if it's a simple Hello World.php, or mysql/php platform, like a forum for example.
Thank you!
Comments
Judging by the Idle CPU, they're overcrowding that server quite alot. Does the server feel sluggish when running commands, and what is the sequential I/O that is usually a good factor.
If the CPU is not powerful enough, it normally brings the I/O down with it.
Thank you Jacob, well the company is RocketVPS (KnownHost). No the server doesn't feel sluggish while typing commands. I was going to benchmark it via ServerBear but can't as the server is now used for production.
Also, the averages do no justice as we have about 4 - 6 concurrent visitors online now, verses 30 or so during peak. It'll jump to 2 - 3, 1 minute average often/randomly. 5 and 15 minute averages most of the time look normal but 1 minute average will jump.
A simple 10 second refresh will throw the CPU to 100% and error out (even with a simple "Hello World").
This looks normal to me, the load is not high for a quadcore system and 0.1% wa doesn't look bad. If the load spikes every time I think it should be the CPU being busy or throttled.
A question off-topic, do you purchase directly from EdgeCast or some resellers?
No, CPU power doesn't affect I/O, bad I/O performance affects CPU load however.
@Noxter
You should run
ioping /dev/sda
+1 for ioping recommendation.
Try that randomly and especially during times you see these perceived slow downs.
Performance from the graphic doesn't look like anything that would worry me.
Other than MySQL everything is under 1% of CPU use.
Load averages are rather low also.
This is absolutely not acceptable, maybe the node is heavily abused causing the CPU power left to you gets really low. Try ask support first, their support should be able to help you with this.
That's mighty strange.
That would seem to be application / application server level issue. Possibly machine resources, many things are though.
Hello world should never timeout with error
It is strange...
Edit: too messy, here it is via pastebin: http://pastebin.com/JTCMV1Cs
http://i.imgur.com/WxkF0lQ.png
http://i.imgur.com/If19HRu.png
My processes don't justify the CPU load. First screenie, my processes are using 10.9%, not it's collective 41.3% on all cores.
Thank you for all the help.
@Fan
GoGrid, $0.3 per GB on their small HTTP platform -- $0.25 per GB on their large platform. $0.60 per GB on their local storage.
P.S: Not sure why the font went bananas.
What kind of storage is that? RAID10?
Because Markdown is a pain in the a....
@Freek
Yup. From what I read, RE4's in RAID10.
Here's another ioping with different data rates:
http://i.imgur.com/6BnOjNa.png