Hi,

I have a reseller account and the performance is somewhere between fair and poor. Sometimes my sites take forever to load, apache seems to go down all the time, and just overall a mediocre server.
I'm really thinking about moving this time (although it's going to be a major pain in the ***).

My question is, after looking around with SSH I noticed I share a server with about 1700 accounts (seems kind of high). I don't have permission to see 'dmesg' so it's hard for me to figure out what exactly it 's running on. I assume its Linux with 1gb ram and dual cpus:

uname -a

Linux xxxx.com 2.4.26-ow3 #1 SMP Fri Aug 6 19:40:40 GMT 2004 i686 unknown

Any ideas on how I best can find out if the server is overloaded?

Type top and paste the output here.

I don't have root so this is what I get:


6:32pm up 15 days, 14:18, 4 users, load average: 321.08, 304.79, 244.83
2 processes: 1 sleeping, 1 running, 0 zombie, 0 stopped
CPU0 states: 28.3% user, 23.1% system, 0.0% nice, 47.4% idle
CPU1 states: 30.0% user, 23.3% system, 0.0% nice, 46.0% idle
Mem: 1033776K av, 1000088K used, 33688K free, 0K shrd, 77316K buff
Swap: 2048248K av, 210468K used, 1837780K free 235064K cached

From what you've described, it doesn't matter if the server is overloaded, poorly maintained, or whatever else the cause may be - if you aren't happy with performance and stability, then you need to move. If your host has another server, they may do this for you, alternatively you'll need to go shopping for another provider (I know, a pain).

In SSH, try running 'top' - this will give you some information about the uptime and load. Alternatively, check the Service Status in cPanel.

I hope this helps - and good luck :)

EDIT - someone beat me to it, but "load average: 321.08, 304.79, 244.83" is very bad. I assume it isn't always like this, so maybe it's a blip - contact your host to see if they are aware of this...

Ask them to move your accounts to a different server with the load average less than 3 or 4. That server is extremely overloaded.

Originally posted by inteltechs
Ask them to move your accounts to a different server with the load average less than 3 or 4. That server is extremely overloaded.

When you say 'load average of less than 3 or 4' and my 'top' output said:

load average: 321.08, 304.79, 244.83

Do you mean those three numbers should be in the single digits as opposed to what I have, triple digits?

That's crazy. That's very overloaded. I'm surprised it's still responding to ssh ;)

It depends on the specs of the server however some can handle 5-10 nicely and some can't even handle 1. If they are at 300 then yes it's overloaded and not by a small margin. Ask them to switch you to a new server or change hosts.

Just because a machine is hitting loads like that now and is overloaded with processes at the current time, does not mean that it is oversubscribed or loaded with too many customers. It would be some feat to actually load up a server so it had consistent loads of 300+.

Whenever we've had issues like this, it's usually someone exploiting a vulnerable formmail or something along those lines. Once that is cleared up, the loads should drop to a more normal level.

This isn't to say that the machine isn't packed to the brim with customers. I'm just saying that there is not enough information here to make that determination.

It's not fair to say the server is overloaded based on a single incident - this could just as easily be a runaway script on the only other account on the server [unlikely, but possible]. The best thing to do is notify your host of the problem, just to make sure they are aware.

If you have noticed poor performance over a longer period of time however, as your original post suggests, then it doesn't matter what the load average is, you'll want to move. :)

Ok thanks guys.

I've been with this company for over a year and have requested to be moved to a different server twice already (first request was like a year ago). But everytime I asked, they said they were putting in a new kernel or tweaking the server which would help the server with performance. It would make a difference for a week and then the same crappy performance would emerge again.

You're probably wondering why I stayed with this host for so long? I had a lot of other things going on my life, school, work, etc. that I never really had the time to move all my sites.

This is the first and only reseller account I've been on and this company was highly recommended on Webhostingtalk a year and half ago. I 'only' paid $25/month and I wasn't sure if other hosts would be better.

With the prices of servers having come down so much since I first got this account, I'll probably get my own budget server from 1and1 or layeredtech.

I'll be sure to out this host once my move is complete (to avoid any retaliation).

Originally posted by Andrew
This isn't to say that the machine isn't packed to the brim with customers. I'm just saying that there is not enough information here to make that determination.

Like I said in my first post, there's 1700 accounts on that machine (i.e. 1700 users in /home)

When those guys told me to run 'top' I did maybe it was a coincidence that I typed 'top' when it was just going crazy, I don't know. But I have had to wait as long as 30 seconds for my page to render. and especially when I'm working on a site I sometimes feel like I personally can make apache crash just by refreshing my browser too many times.

But now I feel pretty confident that they have overloaded the server with accounts.

Thanks to everybody for the advice.

I noticed this on a site, the server goes down for 3 sec's and
then be working fine when you refresh some days this happens
every 10 mins or so...

Whats the problem here

Hi Jazz

Ideally you want something more than 500 actual accounts / websites on a server depending obviously on the server specifications and the types of websites

Not all websites are as small as that though so its a good idea to check with whois.sc using a reverse IP to see how many domains are on the server too as 1700 accounts could mean 2000+ websites??

Originally posted by WESHUK
Hi Jazz

Ideally you want something more than 500 actual accounts / websites on a server depending obviously on the server specifications and the types of websites

Not all websites are as small as that though so its a good idea to check with whois.sc using a reverse IP to see how many domains are on the server too as 1700 accounts could mean 2000+ websites??

For 500 users what would you think the server spec's would
need to be

Well I would not use a server with anything less than 1GB Memory and P4 2.4GHZ if all the websites were very low usage or very small basic websites, or if there are less than 100 websites on the server.

If you have websites using databases then you should look towards dual processor server with 2GB Memory minimum but thats just how I run my business... some of my sites are migrating over to a new dual ZEON CPU server running RAID5 with 2GB memory this week as we have just gone over the 50 websites on 1 server mark but thats just us... overkill

Why not PM me the IP address of your server or simply visit www.whois.sc and create a free account then do a reverse IP lookup on your servers IP address and you will see how many domains point at your server!!!

Let me know how you get on...

WOWOWOW????? I never saw a machine with a load average of 100 respond to SSH within a period of time where you wont time out in. I personally like to keep all my boxes below .60 and at a max of .85 or I am personally not happy with the performance.

load average: 321.08, 304.79, 244.83

good googly moogly! That it even responded to your commands at all is a total miracle.

210468K used

Lotta swapping going on there too, that'll kick a server in the tail when it has to swap to the harddisk. The available memory leads me to the conclusion that the server uses shared video ram, so it's probably not running on a high end server type motherboard (hey nothing "wrong" with that, I have used quite a few shared mem servers in the past and probably will again). What that tells me is this server probably isn't capable of handling 1700 sites even if they were teeny tiny little sites.

When a single cpu server of mine starts showing average cpu usage (5 minute tests via snmp) of over about .6 or a dual over 1.1 or 1.2 then I start getting worried.

It IS possible you just hit it at the wrong minute, however the 15 minute load average is over 200 so this doesn't look like a short load spike, if it were short spike, the 15 minute load average would be like not that much.

Run "uptime" on a cycle, if you have cron privilieges make a 5 minute cron something like this:

*/5 * * * * uptime >> ~/uptime.txt

You can then look at uptime.txt for a rundown of load every 5 minute for a period to see how it sorta "averages" out.

That is really high load! lol

Originally posted by Dixiesys
When a single cpu server of mine starts showing average cpu usage (5 minute tests via snmp) of over about .6 or a dual over 1.1 or 1.2 then I start getting worried.
QUOTE]

[QUOTE]Originally posted by w33t
WOWOWOW????? I never saw a machine with a load average of 100 respond to SSH within a period of time where you wont time out in. I personally like to keep all my boxes below .60 and at a max of .85 or I am personally not happy with the performance.


Great minds think a like :)

I would like to create a cron job set up to write out the date and uptime to a file at 5 min intervals

DO you help me?

Thanks Alot !!!!

AH!? load average: 321.08, 304.79, 244.83 !?
Time to switch server....