How much can a VPS do....

[pkearney]

Hi Guys,

I'm trying to get a handle on what I could/should expect a VPS Hosting to be able to do comfortably. I'd like to have a little survey of what VPS Hostings forum users have, and how much of a load they are putting on them, and whether they experience any memory problems.

I'll start the ball rolling.

I have a VPS Hosting-01.
I've used less than 5Gb of my 10Gb allowance
Monthly bandwidth usage in July was just under 1.5Gb (less than 1% of my possible 200Gb).
Server load rarely goes above 0.1.
Memory usage hovers around the 80% (although when I check memory usage in Main >> Server Status >> CPU/Memory/MySQL Usage the total memory for all the lines listed rarely goes above 20%)
There are 49 accounts on it (some of these accounts would have add-on domains as well, most of which alias to that account).
2 of these accounts are using Fantastico PHPNuke website hostings (both are new and have little to no visitors). 1 Has Fantastico Joomla (this has just been installed it hasn't been setup yet). There are 15 static website hosting and 33 'coming soon' single page website hostings.
There are a total of 69 pop accounts across all the domains and 164 forwarders.
13 of the accounts have Spam Assassin enabled.

My problem is that I have on a number of occasions had to contact support as things haven't worked, and I've been told I'm running out of memory. eUKhosts have added a further 128Mb to my 256Mb of memory I get with the VPS Hosting (which is burstable to 1Gb), but this doesn't seem to have made much difference. For this reason I'm not keen on upgrading until I know what is actually chewing up all the memory.

So where should I be looking to clean things up? What troubleshooting pointers can anyone offer.

Regards

Paul

[DPS Computing]

Unfortunatly, not being a VPS Hosting customer I can't offer such detailed advice but I think you will find what I have to say useful .

You should expect more from a VPS Hosting than shared hosting but a lot less than you should expect from a dedicated server. A dedicated server has much more resources at your disposal and there all for you - not shared throughout other users.

Also, don't forget that all but one customer could be behaving correctly and the server still go down - all it takes is one wayward script on one account that is malfunctioning or being abusive to bring the entire server down. Basically you have got to be able to trust your neighbours on your server.

Maybe eUKhost will be able to get a group of "trusted" VPS Hosting customers to go on a server together - that way the effect of abuse would be minimal?

Hope that helps .

[flesso]

Have you actually checked to see how much memory you are using?

If not, run 'free -m' (without quotes), and that will tell you.

To look at whats chewing up the memory run 'top' (without the quots).

I would copy and paste your results here so that us community members can help you with your problem, you may want to add them to your support ticket as well.

[DPS Computing]

Originally Posted by flesso

Have you actually checked to see how much memory you are using?

If not, run 'free -m' (without quotes), and that will tell you.

To look at whats chewing up the memory run 'top' (without the quots).

I would copy and paste your results here so that us community members can help you with your problem, you may want to add them to your support ticket as well.

That is a good idea flesso - unfortunately if the problem is with another users account you won't be able to see it though - for that you will have to let eUKhost investigate it .

[flesso]

I would also try looking at the amount of incoming mail, as SpamAssassin scans it all then it could be that SpamAssassin is slowly eating away the memory.

DPS, whose to say it's an account? It may be a bad process.

[pkearney]

Hi Flesso,

Thanks for the response. I'm fairly new to Linux, so I'm guessing I need to run these commands from a secure shell using root? (I've used WinSCP3 in the past but just tried to access using it again and I can't authenticate, so I've sent a message to VPS Hosting support).

In a recent note from a VPS Hosting SysAdmin they told me I was using 110Mb of 384Mb. Although I don't understand how the memory allocation works, I'd have thought you should be using pretty much all your memory at any one time, and paging out less active stuff to your page file. Another thing the SysAdmin told me was that I'd had 9136825 fail counts for memory.

I'll run those commands and get the info out here for all to see

On the inbound mail side, I can't say for sure how much mail it is having to scan, but it is weeding out a lot of Spam, so I'm sure there is some load there (I'd say a few hundred spam messages a day across the VPS Hosting). Would it be more on quantity of messages than actual message size (e.g. a message with a large attachment would take a lot longer to scan that 100 small messages??). Is there an (easy - remember I'm pretty new to this )way I can see what the mail traffic has been for a period of time? Is there some sort of counter I could setup so I could monitor this.

Regards

Paul

[flesso]

Originally Posted by pkearney

Hi Flesso,

Thanks for the response. I'm fairly new to Linux, so I'm guessing I need to run these commands from a secure shell using root? (I've used WinSCP3 in the past but just tried to access using it again and I can't authenticate, so I've sent a message to VPS Hosting support).

In a recent note from a VPS Hosting SysAdmin they told me I was using 110Mb of 384Mb. Although I don't understand how the memory allocation works, I'd have thought you should be using pretty much all your memory at any one time, and paging out less active stuff to your page file. Another thing the SysAdmin told me was that I'd had 9136825 fail counts for memory.

I'll run those commands and get the info out here for all to see

Regards

Paul

I would use Putty to run them via SSH.

Once you have run them, paste the results here.

VPS Hosting memory allocations works just like it would on your desktop. You get allocated a gaurunteed amount of memory, in your case 384mb. The burst amount if the amount is the amount which your VPS Hosting can bust up to (if some big request happens for a few seconds), this may not always be available as others on the node will be used the burst space as well.

In Short:

Guaranteed: For Your Use Only
Burst: Shared With Others

[DPS Computing]

Originally Posted by flesso

I would also try looking at the amount of incoming mail, as SpamAssassin scans it all then it could be that SpamAssassin is slowly eating away the memory.

DPS, whose to say it's an account? It may be a bad process.

Oh I am not saying it is - just saying that this is a common problem that sometimes causes VPS Hosting failure. You are of course right - it could be a bad process.

[pkearney]

OK Here goes

I've logged in and run the 'top' command and this is what is displayed. It doesn't mean a lot to me but I don't know why it is saying 200Mb for memory (when it should be 384MB). The free -m is at the end of the list. Let me know if there is more info you need, or if I need to copy something else from PuTTY (I have only just worked out that you just highlight what you'd like to copy and then click paste here).

Regards

Paul


top - 16:19:45 up 15:26, 1 user, load average: 0.09, 0.06, 0.01
Tasks: 75 total, 1 running, 74 sleeping, 0 stopped, 0 zombie
Cpu(s): 0.5% us, 0.3% sy, 0.0% ni, 99.2% id, 0.0% wa, 0.0% hi, 0.0% si
Mem: 2068072k total, 2048360k used, 19712k free, 45240k buffers
Swap: 2096440k total, 885188k used, 1211252k free, 407372k cached

PID USER Page Ranking NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
3500 nobody 16 0 102m 90m 2652 S 0 4.5 0:35.45 httpd
15454 root 16 0 35692 30m 2048 S 0 1.5 0:17.46 spamd
15455 root 16 0 33436 28m 2016 S 0 1.4 0:01.18 spamd
23703 root 16 0 30992 25m 1876 S 0 1.3 0:02.84 spamd
5281 nobody 15 0 31740 19m 2796 S 0 0.9 0:20.53 httpd
3499 nobody 17 0 30792 18m 3276 S 0 0.9 0:24.21 httpd
24521 nobody 17 0 31020 18m 2756 S 0 0.9 0:48.06 httpd
3498 nobody 16 0 30912 18m 2800 S 0 0.9 0:22.54 httpd
24509 nobody 17 0 30896 18m 2776 S 0 0.9 0:09.30 httpd
3994 nobody 15 0 30868 18m 2796 S 0 0.9 0:48.41 httpd
3496 nobody 15 0 30756 18m 2800 S 0 0.9 0:49.00 httpd
24531 nobody 16 0 30556 17m 2680 S 0 0.9 0:16.80 httpd
3497 nobody 15 0 30332 17m 2796 S 0 0.9 0:51.68 httpd
12089 nobody 16 0 30348 17m 2776 S 0 0.9 0:19.03 httpd
1567 MySQL Server 16 0 105m 17m 3900 S 0 0.9 0:19.96 mysqld
3384 root 16 0 22740 13m 3792 S 0 0.7 0:00.93 httpd
4089 root 16 0 10196 8464 4288 S 0 0.4 0:01.90 authProg
5498 root 17 0 10196 8464 4288 S 0 0.4 0:02.16 authProg
25704 root 17 0 10196 8464 4288 S 0 0.4 0:02.51 authProg
26505 root 16 0 10196 8464 4288 S 0 0.4 0:02.17 authProg
31905 root 16 0 10196 8464 4288 S 0 0.4 0:01.54 authProg
1607 root 15 0 11936 8088 1092 S 0 0.4 0:01.11 chkservd
3603 root 16 0 16744 7880 1008 S 0 0.4 0:03.08 cpsrvd-ssl
3560 root 39 19 11232 7604 1692 S 0 0.4 0:13.63 cpanellogd
3746 mailman 15 0 11324 6752 2604 S 0 0.3 0:01.09 python2.4
3745 mailman 15 0 11648 6708 2604 S 0 0.3 0:01.01 python2.4
3742 mailman 15 0 10792 6692 2604 S 0 0.3 0:01.01 python2.4
3743 mailman 15 0 11300 6692 2604 S 0 0.3 0:01.01 python2.4
3741 mailman 15 0 10364 6684 2604 S 0 0.3 0:01.04 python2.4
3744 mailman 15 0 11528 6684 2604 S 0 0.3 0:01.00 python2.4
3747 mailman 15 0 11636 6680 2604 S 0 0.3 0:00.95 python2.4
3748 mailman 16 0 12068 6680 2604 S 0 0.3 0:00.13 python2.4
3507 mailnull 15 0 9836 5540 1544 S 0 0.3 0:00.99 eximstats
3559 root 15 0 8584 5248 1072 S 0 0.3 0:00.88 cpbandwd
3705 mailman 19 0 10768 4920 828 S 0 0.2 0:00.00 mailmanctl
30563 named 21 0 48604 3772 1964 S 0 0.2 0:02.77 named
28569 root 15 0 8208 2284 1832 S 0 0.1 0:00.02 sshd
31826 root 15 0 7340 2264 1828 S 0 0.1 0:00.01 sshd
32186 root 15 0 6256 2128 1724 S 0 0.1 0:00.01 sshd

root@tiger [~]# free -m
total used free shared buffers cached
Mem: 2019 1965 54 0 39 366
-/+ buffers/cache: 1559 460
Swap: 2047 864 1182

[flesso]

Well, the processes look ok as only 1 is running, and the other 74 are sleeping.

Your memory looks fine as well, you have a good amount free.

I would have a go at tweaking cPanel as I have discussed in this thread: Tweaking cPanel

That should reduce your overall cpu and memory footprint, and should be ok for you as you don't use many resources anyway.

[DPS Computing]

If you don't need things like SpamAssassin disable them through cPanel - they are very resource intensive - especially if you get a lot of email.

Only use the features that you want to us and don't leave things enabled for the sake of leaving it enabled - as if everyone did this on the VPS Hosting it is easy to understand how the server can quickly and without warning become overloaded

[pkearney]

Hi,

Thanks Flesso/DPS.

I'm confused about the memory numbers though. Just had a look at my Dedicated Server status and have this, all green lighted, but the memory used is just under 80% which is when it goes to a warning amber light.

spamd up
imap up
cpsrvd up
exim (exim-4.63-1_cpanel_maildir) up
mysql (4.1.21-standard) up
pop up
named (9.2.4) up
httpd (1.3.37 (Unix)) up
ftpd up
Server Load 0.02 (2 cpus)
Memory Used 77.9 %
Swap Used 42.21 %
Disk simfs (/) 50 %

But how do these number relate to what we see in top?
Mem: 2068072k total, 2048360k used, 19712k free, 45240k buffers
Swap: 2096440k total, 885188k used, 1211252k free, 407372k cached

Where it seems I actually only have 19Mb free from 2Gb??? Or am I reading thiese numbers wrong?.

I can't stop Spam Assassin as it is working well for us and catching hundreds of spam messages a day, but I have limited it to being enabled just on the accounts that are getting heavy spam (13 out of 49).

I was planning on using Mailman on one of our accounts to manage a mailing list, but if it is a resource hog then I'll look at stopping it.

I've set the catchall's to fail and removed squirrel mail, box trapper spam trap, and spam assassin spam box delivery to see if they make any difference. I have created a custom pop acount that filters all mail for an account that is marked with a spam score of 5 or more so the users can whitelist any incorrectly tagged messages. Eventually I hope to set it to just delete the spam messages once the respective users are convinced nothing genuine will get tagged.

I'll do the Apache changes after I've seen what sort of affect these changes have.

But the whole memory thing is still confusing me. I'd expect to see somewhere that I've the 384Mb my VPS Hosting has, and a percentage of that in use (that corrolates with the percentage in use we see in the server status), but the number just don't add up...

Thanks again for your help with this.

Regards

Paul

[eUKShane]

Hey,

First of all, you cannot check memory usage using "free -m" command on VPS Hosting's. As it shows the usage of the Hardware Node on which your VPS Hosting is currently hosted.

You can check assigned memory Or used memory using "cat /proc/user_beancounters" command.

Check privvmpages (barrier) limit, in your case it should be 393216 (384 MB).

for more detailed information please refer : http://forum.swsoft.com/showthread.p...threadid=26770

For further assistance, please email to <a href="http://www.eukhost.com/vps-...a>@eukhost.com. So that we will explain you in a better way.

[DPS Computing]

If SpamAssassin is catching hundreds of messages a day for you then it will be an intensivly running process so just make a sure you keep your eye on it .

[pkearney]

Hi,

Whhooaaahhh!!! OK, it is going to take me some time to understand that forum link, but in the meantime these are the numbers I get when I run that command.

root@tiger [~]# cat /proc/user_beancounters
Version: 2.5
uid resource held maxheld barrier limit failcnt
123: kmemsize 9522522 22491483 41943040 44711280 0
lockedpages 0 0 128 128 0
privvmpages 145351 274549 262144 393216 9136825
shmpages 1371 3115 8192 8192 0
dummy 0 0 0 0 0
numproc 83 222 500 500 0
physpages 88649 183981 0 2147483647 0
vmguarpages 0 0 262144 2147483647 0
oomguarpages 89805 212711 65536 2147483647 0
numtcpsock 28 79 800 800 0
numflock 9 37 100 110 0
numpty 2 2 16 16 0
numsiginfo 0 77 256 256 0
tcpsndbuf 254904 630552 3194880 5242880 0
tcprcvbuf 396152 761768 3194880 5242880 0
othersockbuf 40248 676272 1320960 3368960 0
dgramrcvbuf 0 130556 1320960 1320960 0
numothersock 34 74 800 800 0
dcachesize 0 0 5242880 5488640 0
numfile 2716 5614 12800 12800 0
dummy 0 0 0 0 0
dummy 0 0 0 0 0
dummy 0 0 0 0 0
numiptent 276 280 500 500 0

Now all I can really see from this is that I have a really high failcount on privvmpages. But I can't see what is causing it? Now according to the top command the node has 2Gb of memory, but how many VPS Hostings is this shared between? How does the guaranteed memory allocation work (e.g.you could only 'fit' 8 VPS Hosting-01s on a 2Gb server if they all had 256Mb guaranteed, and that doesn't leave any space for 'bursting' unless you are bursting into the other users guaranteed space).

You have already increased my memory for me from 256 to 384, but it doesn't seem to have made any difference. Is there a way of resetting the counters for this file so we can se if there are any new failcounts in case it is just showing us old ones?

What is actually happening when a failcount occurs? Can processes get hung if this happens, there by not releasing their memory back to the VPS Hosting? In my earlier list of everything in the 'top' command there were 75 processes of which 74 were sleeping, but the server was still showing a memory usage fo just under 80%?

We are getting there, but I still can't really see how I can diagnose what is using all the memory when I go above 80%.

Regards

Paul