Re: Interpretation of top statistics

From: TNEICSEN (veevii_at_gmail.com)
Date: 11/05/04 

Date: 4 Nov 2004 20:21:08 -0800

You need to look more I guess. No idea what applications you run, what
OS version... 2 cents below

>From your vmstat r column run queue is high. This means the processor
queue is building up. How many CPUs do you have ? This is normally <=
NoofCPUs x 2 on a non stressed system. Could be because the processor
is loaded fully (probably not your case) or the system is doing
something else like I/O when the processor is blocked.

>From your vmstat w column the IO wait (or Buffer wait) seems high. Are
you applications I/O intensive ?

Try the swapinfo -tm to study the mem usage including swap. See what %
is reported as free when the freeze happens.

Do you get any errors like no memory to fork or not enough space or
some such error which indicates a memory or process or thread limit
being hit ? Look at syslog and also application logs for any OS
returned errors.

Look for other commands like sar/iostat to study disk I/O

Look at netstat -an to see if sockets are being open too frequently
and not closed.

See ipcs -am to look if shared memory is building for your
applications

***************************************************************************

jerolain@free.fr (Jerolain) wrote in message news:<3ef82f60.0411020810.40dc22e5@posting.google.com>...
> The problem I'm trying to track is that every now and then, the system
> is freezing. For instance, users using a telnet session, experience a
> long delay in the echo of the character they strike.
>
> I have the top command running on a server, reporting servers
> statistics every 5 second.
>
> Here is what top reported at a time a freeze of the system :
>
> System: stfo001 Tue Aug 3 11:38:37 2004
> Load averages: 7.10, 7.64, 8.44
> 276 processes: 222 sleeping, 52 running, 2 zombies
> Cpu states:
> LOAD USER NICE SYS IDLE BLOCK SWAIT INTR SSYS
> 7.10 1.0% 0.0% 2.2% 96.8% 0.0% 0.0% 0.0% 0.0%
>
> Memory: 796180K (311204K) real, 4755244K (4019668K) virtual, 32968K
> free Page# 1/6
>
>
> System: stfo001 Tue Aug 3 11:38:55 2004
> Load averages: 7.72, 7.74, 8.46
> 276 processes: 232 sleeping, 44 running
> Cpu states:
> LOAD USER NICE SYS IDLE BLOCK SWAIT INTR SSYS
> 7.72 6.4% 0.0% 2.9% 90.7% 0.0% 0.0% 0.0% 0.0%
>
> Memory: 805964K (322720K) real, 4767420K (4055116K) virtual, 31248K
> free Page# 1/6
>
> So there are 18 seconds between the two statistics instead of 5. First
> obvious conclusion is that the system is really freezed :)
>
> But what can I conclude from the top information or what should I look
> ? Are the statistics reliable as they should be produced every 5
> seconds and some are missing ? (for instance CPU time is not at all
> 100% percent).
>
> I have in parallel vmstat running every 5s and here are the statistics
> reported at the same time than top :
>
> 03113715: procs memory page faults cpu
> 03113715: r b w avm free re at pi po fr de sr in sy cs us sy id
> ...
> 03113826: 11 1 9 987875 8644 157 39 227 6 13 0 48 325 8441 878 11 12
> 76
> 03113849: 11 0 32 1004917 7852 26 2 28 0 0 0 0 246 1580 701 3 2 95
>
> I see also a delay between the two stats (here 23s). The po in the
> first line is 27 and the sr is 48, whih could be interpreted as quite
> high .
>
> My first guess is that the system lacks some memory. Could the system
> swap explain such a freeze ? The users usuallay report a system hang
> during aroung 40s to 1mn. It's quite impresssive for only a system
> swap.
>
> In that case, I have a question concerning the memory usage as related
> by top. I can see the physical memory usage reported by top is 805964K
> which is quite near the total memory (1GB). Is this usage as reported
> by top accurate ? And what is the number in parenthesis. The man page
> says it is memory considered as "active". I don't get the meaning.
>
> Well that's a lot of questions :)
>
> PS : I'm not the system administrator and I don't want to be, I am
> just here to diagnose this freezing problem
>
> Thanks for any help :)

Relevant Pages

  • Re: Slowly Terminal Server response
    ... physical memory is an indicator of available memory. ... If Task Manager only shows a single CPU graph, then only one CPU is installed. ... A good general guide to keep in mind is that most applications require the similar amount of memory as they would on a desktop for each user. ... I normally start with a general guideline of 25-30 concurrent users for a server with 4 GB of RAM and 2 CPUs and then do testing for a specific set of required applications. ...
    (microsoft.public.windows.terminal_services)
  • VIA C7 / VIA PC-1 (PC2500) anyone?
    ... fanless, -- it becomes a almost real PC (1500MHz CPU), ... Segmentation fault ... It's definitely NOT memory issue - I tried several different ... -- just random applications are ...
    (Linux-Kernel)
  • Re: Intel processor at wrong speed
    ... freezing and sluggish, and after running Belarc, I noted that the ... processor was reporting as a 2.13 rather than 3.20. ... I have a couple months of warranty left on the CPU and am ... memory while testing. ...
    (microsoft.public.windowsxp.hardware)
  • Questions about Java SE Real-Time
    ... party applications can end up on the same device, ... One problem in this context is the arbitration of memory and CPU among ... Can Java SE Real-Time set an arbitrary limits to the CPU ...
    (comp.lang.java.programmer)
  • Re: 64bit CPU -or- dual cpu m/b ?
    ... > 64 bitness has to do with the address space of the CPU not with the type ... > high bandwidth link between CPUs in a multiprocessor system, ... > address space gives a huge performance advantage to applications that need ... > the Opteron is the best choice because each CPU has it's own memory system ...
    (alt.os.linux)