service failures, massive server loads

[superbunker]

For the last few months, I've been getting repeated services failures on my VPS. Mostly ftpd, but sometimes exim and httpd. When this happens (and still I'm able to get into WHM) the service status shows very high loads -- today when this happened, I had loads above 9.3 for a while.

I've opened several tickets about this issue, which usually conclude with "looks normal now" or something equally helpful. I have gotten confirmation that there is a problem with the hardware node. Once, I was told it was another user on the node, and that the user had been removed. And yet today, I lost cpanel, ftp and email for a good while.

The current ticket I have open on this has been open for a month. I'm not sure what to do, since this seems to be out of the ordinary for JaguarPC, but support is clearly unable to deal with the issue. I don't want to go through the hell of switching servers, but this unstable VPS account is becoming a serious problem for me and my users...

Does anyone have any advice?

[mattsiegman]

Ask for the ticket to be given to Masood, or PM Greg on here.

[thisisit3]

Is your VPS generating the cpu load or another VPS? In theory you are given an X amount of cpu usage as a minimum and another VPS shouldn't generate such a load on yours. Maybe i'm wrong...

[JPC-Veena]

superbunker,

Please pm me or send here your recent ticket number and I shall check where the problem lies.

[the_ancient]

, I was told it was another user on the node, and that the user had been removed.

That contradicts the main selling point of a VPS


Damn L1 Tech

[superbunker]

Is your VPS generating the cpu load or another VPS? In theory you are given an X amount of cpu usage as a minimum and another VPS shouldn't generate such a load on yours. Maybe i'm wrong...

I asked several times if my VPS was causing this -- I was told that my account isn't generating the load. Once I was told it was another user on the node, this time I was just told it was the hardware node.

I also thought that having a set level of resources was the point of a VPS. Just as important as getting this fixed, I need to know the facts about this issue.

The current open ticket is #2460508. (Several of my past tickets are about this too)

[JPC-Veena]

I checked your ticket and your vps. The techs have informed you about the resources your vps was hitting and have already increased them for you too. You can see them under your vz powerpanel -> qos alerts section.
I can also assure you no service failure on your vps can be caused by anything which happens outside of your vps, so you need to monitor your vps closely. Also, the guaranteed resources of your vps are always available to you irrespective of external factors. The only exception would be cpu, we have an equal share cpu environment and if any user bursts above the allowed limit, we notify them to upgrade to a dedicated server.

[superbunker]

Veena,

Here's ticket 2451014:

- - - Tech - - - : Smith. 2007-02-12 / 17:48:03
I notice a VPS that is causing high load issues on the entire Node and I have advised him to upgrade to a dedicated server otherwise we have to suspend his VPS.The reason that other users can effect how your VPS performs is because we run equal cpu sharing on the VPS. This means that you have a guaranteed amount of cpu for your VPS and you are allowed to burst to the full amount of the cpu if it it available. The problem is when some users begin bursting to the full amount of the cpu which leaves you with only the guaranteed amount of cpu that you have. Since you are not allowed to burst above your guaranteed cpu usage your performance is effected.

Let us know if you need further assistance.

Regards,
Smith

+ + + Client + + + : Jason.Brown 2007-02-12 / 17:57:35

Just to be clear --
it was another VPS on my node and not my VPS? Am I okay?

- - - SYSTEM - - - 2007-02-12 / 17:58:58
A customer service representative has begun working on this issue.

- - - Tech - - - : Smith. 2007-02-12 / 18:00:57
Yes.

Let us know if you need further assistance.

Regards,
Smith

[JPC-Veena]

Yes, that was over a month back, I am referring to current situation.

[superbunker]

If something on my VPS is actually causing this I really need to know what it is. I don't run anything that should be using the entire CPU.

[superbunker]

Yes, that was over a month back, I am referring to current situation.

Currently open ticket, reply from last night:

I have checked there was some load on the hardware node and I have monitored that.

Can you explain what that means?

[JPC-Veena]

Sorry, you could have asked the tech who said that. In any case, I was not saying your vps is causing cpu issues, I just said the problem you are seeing is ftp restarts and they are coinciding with vz parameters hitting limits on your vps, numtcpsock and another one. If you see a restart message, you need to login to your vps at the time and run "top" and sort by cpu/memory and see what processes are causing the limits to be hit and fix that. You can also contact us at the time so we can see what it is while it is happening.
About the hardware node, all issues on the hardware node are taken care of by us.
For any further concerns, please update your support ticket.

[superbunker]

You can also contact us at the time so we can see what it is while it is happening.
About the hardware node, all issues on the hardware node are taken care of by us.
For any further concerns, please update your support ticket.

Thanks, Veena. But the reason I'm here is because I have been doing that through support tickets for the last few months. I open a ticket when it's happening, and the response I've gotten from the support techs is that it is not anything on my VPS. When I ask for more explanation, I'm told "Please check now."

At this point, I'm getting seriously fed up.

[superbunker]

Can someone just clarify a basic issue for me --

Can other users on my node cause service failures on my VPS?
So far, the answers I've gotten are "yes", "no" and "maybe."

[thisisit3]

In theory they can't but apparently in practice they can!

I'm also having problems with QoS limits, the default limits are set WAYYY TOO LOW, even simple things with PHP's GD functions or ImageMagick will hit them really fast. My biggest problem is "shmpages".

Here is a little tip:

# cat /proc/user_beancounters
...

That should help you monitor your resources and QoS limits much better than the VZPP QoS page.