Hi, folks-

Just a quick question - out of curiousity, I did an "uptime" while telnetted into my host's server... and this is what I got:

11:24pm up 9:51, 7 users, load average: 8.79, 6.67, 4.75

I'm not super clear on this, but are the load averages MAJORLY high, or is it just me?

Thanks for the help!

Neil

Yes the load is a bit high. This would only be a problem if its always at this level, it wouldnt be unusual for server loads to rise during peak processing times.



TWA

I understand that our server load is very low. LOL, I just cut this line out and it looks like the server is sleeping. It's not always this way, it's just a slow morning.

8:59am up 25 days, 22:26, 1 user, load average: 0.00, 0.00, 0.00

But it's usually looks something more like this:

8:59am up 25 days, 22:26, 1 user, load average: 0.05, 0.00, 0.00

I heard that if the first number exceeds 5.00 then the load is high. How do I read these numbers and understand what they mean?

Originally posted by Farnsworth & Thomas

How do I read these numbers and understand what they mean?

Uhmm..let say we got the following info:

8:59am up 25 days, 22:26, 1 user, load average: 1.00, 1.75, 0.60

Recent time for your server is 8:59.
and the server already "live" 25 days plus 22:26 hours - since the last reboot.
Recently there's 1 user log into shell access.
Load average on average:
Load server on the last 1 minute is 1.00
Load server on the last 5 minutes is 1.75
Load server on the last 15 minutes is 0.60

Hope this help.

Let's take apart this line:

8:59am up 25 days, 22:26, 1 user, load average: 0.05, 0.00, 0.00

First you see the server time, number of days and hours "up", number of users logged in via telnet or SSH. Last we have the load average. If you watch your session, the first number (0.05) usually changes every time your screen refreshes giving you the current load status. The second number is an average over 5 or 10 minutes, the last number is the load average over 10 or 15 minutes.

By looking at the first number it is easy to see what is going on at that point in time. The other two averages can give you a general idea about what has been happening for at least the last 15 minutes, so if your first number was really high (we'll say over 5.0) and the last number was really high there is something going on that should be looked into. If the first number is really high but the last number is fairly low, there may be just some spiking going on which is normal.

In this particular case, you have some activity, but your server has been pretty quiet for at least 15 minutes before getting those numbers. I hope this helps and that I have explained it correctly!

Thanks everyone.

What is considered a bad load average?

what's the unit of those load figures?

It can't be percent can it? If 100% was "CPU running on all cilinders" then 5 or 6% would be low wouldn't it?

What figure would mean full load?

Mine produces this. Yet another command I hadn't heared of before :)

uptime
5:11pm up 3 days, 18:13, 1 user, load average: 0.01, 0.01, 0.00

I'm pretty sure 1.00 is 100%. Although if it is a dual-cpu server 2.00 would be 100% and so on.
You can also find the load info by running "top"

Hope that helps :)

Well low load averages are the best load averages.
My load averages are at http://www.iboxhosting.com/serverinfo

What happens if your load averages are high is the server will become slow.

For example, if your load averages reach 25.0 at any point usually it takes you longer to login throught telnet or ssh, your webpage comes up slower and things like that.

I know on my server we maintain great load averages however there have been times were users would overload servers by running crappy cgi-perl scripts every 2 seconds.

"pstree" is a cool tool also to see current processes and "ps aux".. i usually run "ps -A" for those of you who want to explore with some cool commands.

How do I read this:

$ df
Filesystem 1k-blocks Used Available Use% Mounted on
/dev/hda1 746418 510644 235774 68% /
/dev/hda3 202188 10763 191425 5% /var
/dev/hda4 6882882 1033164 5849718 15% /home

'/home' looks like the root with ~7Gb space(?)
But what is '/' then (with only ~750Mb space) or '/var'?

Thanks

re Load average: the load average is the average number of processes in the run queue. The run queue is the list of processes which are waiting for CPU time. So a load average of 5.0 means that on average you've got 5 processes which want to do stuff. In general, if you double the load average, everything takes twice as long. (Turning technical here, you may want to stop reading) If we make the totally false assumption that processes arriving in the run queue can be modeled as independant memoryless random variables, then the run queue is of length l^2/(1-l) where l is the real load (ie, utilization). A load average of 5.0 would then mean a utilization of 85%.

In general, it is a good idea to keep the load average below 1.0. For most applications this means running at about 60% of peak capacity; however, once you move beyond this point performance degrades noticably.

One exception to this is niced processes. Cron jobs (ie, log parsing) are usually set to run at low priority, so they will not interfere with server operation even though they increase the load average.

to topp: your system has three partitions, / /var and /home. /home is 7GB and contains the home directories for everyone, all your websites, and stuff like that. /var contains data used by server processes: For example, /var/spool/mail is sendmail's mail spool, /var/mail/xyz is a mail spool for user xyz, /var/log/ usually contains log files. / contains everything else: root's home directory in /root (a good reason why you should never do anything as root!), the kernel, system files (in /bin,/sbin,/usr/bin, and /usr/sbin), configuration files (in /etc), and various other miscellany.

Basically, /home is stuff that users mess with, /var is stuff that server processes use, and / is stuff which should only change when you are setting the server up.

Hope that all helps someone :)

Going along with Farnsworth & Thomas on this, on a high end server, a load greater than 5 is when you will start to notice a degradation in service speed.

When you have something like 8.79, 6.67, 4.75 this means that the server is experiencing an abnormally high amount of load during the last few minutes but not consistantly.

-Brendan

Originally posted by cperciva
to topp: your system has three partitions, / /var and /home. /home is 7GB and contains the home directories for everyone, all your websites, and stuff like that. /var contains data used by server processes: For example, /var/spool/mail is sendmail's mail spool, /var/mail/xyz is a mail spool for user xyz, /var/log/ usually contains log files. / contains everything else: root's home directory in /root (a good reason why you should never do anything as root!), the kernel, system files (in /bin,/sbin,/usr/bin, and /usr/sbin), configuration files (in /etc), and various other miscellany.

Basically, /home is stuff that users mess with, /var is stuff that server processes use, and / is stuff which should only change when you are setting the server up.

Hope that all helps someone :)

Thanks cperciva!

Thats TERRIBLE! Your load average should be around .08 - 1.0

Try running "top" and see whats taking up all the mem.


James R. Clark II
Nethosters Inc.
http://www.nethosters.com

Heres a cool thing that Linux does. It calculates the bogomips for a machine. Bogomips are the millions of times per second that a processor can do absolutely nothing ; ) Its almost completely useless, but I just think its cool that there is something like that ; )

At the prompt:
cat /proc/cpuinfo

That will give you the bogomips as well as a lot of other info...