98% Memory Used - All Sites Locked Up

45n5

New Member
This evening i couldn't access any site sitting on my server.

When i went into my WHM panel and went to the Server Status->Service Status it had a red flashing light next to "memory used" and was showing 98.2%.

I did a "graceful reboot" and all of my sites came back up in a few seconds but now cpsrvd is flashing red and says "failed".

What happened? Will my sites really shut off if memory used goes to 98% again?

Most importantly, how do I make sure this won't happen again?

Thanks for the help.
mark
 
Disable unneeded services (Spam Assassin is particularly wasteful) and/or purchase a memory upgrade.
 
I disabled Spam Assassin but it didn't seem to help much.

My server is running at 40% cpu usage consistently. I will be upgrading the ram today.

It still doesn't make any sense that ALL my sites will go down because I didn't purchase a few dollar memory upgrade.

I specifically asked about this in presales and they said I would be notified when I need to upgrade any account features long before they affect my account status. I reckon this is not the case?

Thanks for the suggestions all.
 
This may not be the case for you, but check your memory usage using the "free -m" command from the shell.

My WHM memory usage was showing eg 90% usage and I asked support about it. They told me not to worry about that because checking using "free -m" or "cat /proc/meminfo" showed that the usage was at 75%.
 
thanks all

I went ahead a did a support ticket and they recommended a slight ram uprgrade.

I went ahead and upgraded a +256 to give me plenty of headroom ;)

Too bad they don't warn you ahead of time when your using to many resources. I know to pay better attention to that now.

Do they have a checklist for newbies anywhere of the the critical metrics for running a vps? Like if you could only check these 10 stats everyday, this is where you want to look.
 
It still doesn't make any sense that ALL my sites will go down because I didn't purchase a few dollar memory upgrade.

Your entire car can stop running if you don't purchase $3 worth of gas; this isn't any different. If the resources your server processes need to run are not available then they can't do anything.
 
Your entire car can stop running if you don't purchase $3 worth of gas; this isn't any different. If the resources your server processes need to run are not available then they can't do anything.

A decent car has a warning light to tell me I'm gonna run out of gas long before the situation turns serious. On the other hand, Knownhost just died without warning.

Comparing knownhost to car care doesn't help anybody.

And I'm still a satisfied knownhost customer. Just hoping I can get more warning next time like promised in presales.

I'm still curious if anybody knows

"Do they have a checklist for newbies anywhere of the the critical metrics for running a vps? Like if you could only check these 10 stats everyday, this is where you want to look."
 
Hello,

I wanted to chime in here for a second to make sure things aren't confused. We, KnownHost can't proactively email/call if customers are running out of memory and I don't think any customer expects us to. We do use Virtuozzo as a layer on all our servers which we also include the VZPP (Virtuozzzo Powerpanel) on port 4643 which has a feature that should let you know when your hitting memory limits. This is how we notify you but it isn't thru personal interaction rather thru a product we use and pass down to our clients. If you are seeing these warnings then the best approach is to talk to our Support team to see what can be done to help cut back the memory usage (you may have done this already). If it's not possible to cut back memory usage as you require all applications running then upgrading your VPS is the only answer. So in general the VZPP lets you know if your hitting memory limits but we won't be contacting you if you hit memory limits. I hope this all makes sense.

Thanks,
Joel
 
A decent car has a warning light to tell me I'm gonna run out of gas long before the situation turns serious.

And what happens if the driver simply chooses not to look at that warning light? Then they'd complain that it didn't buzz, and if it buzzed, then they'd complain that it didn't blink, and if it blinked, then they'd complain that it didn't draw them a map to the nearest gas station and buy the gas for them. Some degree of awareness is always required.

If you're using WHM then you have a Service Status page that provides all of the information you want at a glance, including little warning lights. Knownhost will warn you if you're going to exceed your bandwidth limits to a degree that will result in them shutting your account down, but not if you install something that just happens to suck up all of your CPU time or memory and renders a service inoperable. Not everybody needs the same services, and they have no way of knowing what constitutes a failure for you specifically. I kill IMAP on purpose, so I don't want anybody warning me that IMAP is down. Somebody else who relies on IMAP would have a completely different take on the issue, so one size is simply never going to fit all.

It would be fairly trivial to write a script that would check all of these things for you periodically and send you an email whenever a predetermined condition is met, so I'm sure there are some out there that do exactly that. WHM itself will provide notification of a process' death if you check the "Monitor" box next to it in the Service Manager, but it won't tell you how often its hitting the swap file trying to keep them alive.
 
Hello,

We do use Virtuozzo as a layer on all our servers which we also include the VZPP (Virtuozzzo Powerpanel) on port 4643 which has a feature that should let you know when your hitting memory limits. This is how we notify you but it isn't thru personal interaction rather thru a product we use and pass down to our clients.

Thanks joel.

"which has a feature"

What is the name of the feature?

I looked around and I think "QoS Alerts" is what you are referring to? Shouldn't there be an alert there from the other evening or do I need to set them up myself?
 
You should see something. If you know the exact time then this should be verifiable. I would definitely ask Support on this. It is a bit out of my area.

Yes, it is the QoS alerts that I was referring to.

Joel
 
A decent car has a warning light to tell me I'm gonna run out of gas long before the situation turns serious. On the other hand, Knownhost just died without warning.

Why don't you install a program called Load Average, which is a brilliant program that I've been using for years. You can easily monitor memory usage and load averages over a period of time - the best 'warning light' you could get for a VPS.

If you want to look at my loadaverage page, drop me a PM and I'll send you the link (I won't make it publicly known for operational reasons)

http://www.labradordata.ca/home/37
 
Why don't you install a program called Load Average, which is a brilliant program that I've been using for years. You can easily monitor memory usage and load averages over a period of time - the best 'warning light' you could get for a VPS.

If you want to look at my loadaverage page, drop me a PM and I'll send you the link (I won't make it publicly known for operational reasons)

http://www.labradordata.ca/home/37

Im experimenting another one that got many nice review on webhost talks

http://www.nagios.org/

throw me PM if you need to show a demo on my VPS

Interesting one
 
Top