Performance problems...

[Ugo Bellavance]

> -----Message d'origine-----
> De : Eric J Merkel [mailto:merkel at METALINK.NET]
> Envoyé : Friday, January 30, 2004 4:07 PM
> À : MAILSCANNER at JISCMAIL.AC.UK
> Objet : Re: Performance problems...
> 
<snip>

> > >
> > > The specs on the system are as follows:
> > >
> > > Intel PIII 700Mhz
> > > 512K RAM
> > > Ultra2-160 SCSI LVD drives
> > > RedHat 9 + Sendmail 8.12.8 + MailScanner 4.25-14
> > >
> > > I know this system is not very high powered, but I have
> > > systems identical to
> > > this one in a SMTP load-balanced cluster runnig RAV +
> > > libmilter wirh no such
> > > issues.
> > >
> > > Thanks for any advice you might lend.
> >
> > Please let you know the output of :
> >
> > free
> >
> > (to see if you are swapping)
> 
> It doesn't appear to be swapping...
> 
>              total       used       free     shared    
> buffers     cached
> Mem:        514684     352560     162124          0      
> 91748      80912
> -/+ buffers/cache:     179900     334784
> Swap:      1020116        124    1019992

You are right.  But make sure you make top and look if it ever swaps.
> 
> >
> > vmstat 5
> >
> > iostat 5
> >
> 
> I do not seem to have iostat on this machine but here is a 
> sample of the
> vmstat.
> 
>    procs                      memory      swap          io     system
> cpu
>  r  b  w   swpd   free   buff  cache   si   so    bi    bo   
> in    cs us sy
> id
> 10  0  0    124  96680  92392  79592    0    0     0  1045  
> 460   574 54 46
> 0
> 10  2  2    124  73732  92396  86048    0    0  1225  1006  
> 612   851 58 42
> 0
>  8  2  1    124  65824  92396  93740    0    0  1454  1766  
> 807  1114 39 30
> 31
> 15  0  0    124  69712  92396  94212    0    0    34  1318  
> 518   721 62 38
> 0
>  7  0  0    124  81264  92404  94660    0    0     0  1736  
> 602   852 67 33
> 0
> 11  4  1    124  84936  92408  95012    0    0     0  1547  
> 493   763 47 23
> 30
> 13  1  2    124  92588  92412  94440    0    0     0  1803  
> 510   761 63 37
> 0
>  9  0  1    124  84756  92416  94764    0    0     0  1641  
> 584   849 58 42
> 0
>  8  2  1    124  87576  92420  94612    0    0     0  1310  
> 576   773 65 35
> 0
>  8  1  1    124  83644  92424  94956    0    0     0  1365  
> 475   672 68 32
> 0
> 10  0  0    124  81748  92428  94148    0    0     0  1412  
> 463   662 64 36
> 0
> 13  0  0    124  81020  92432  94656    0    0     0  1192  
> 448   664 60 40
> 0
>  9  1  1    124  79820  92436  94968    0    0     0  1106  
> 477   702 57 43
> 0
>  9  4  1    124  73908  92440  95200    0    0     0   859  
> 363   492 50 34
> 16
>  9  1  1    124  76824  92444  95544    0    0     0   894  
> 363   557 63 37
> 0
> 12  0  1    124  89176  92452  95840    0    0     0  1388  
> 545   758 71 29
> 0
>  8  0  0    124  79660  92456  96108    0    0     0  1355  
> 496   708 69 31
> 0
>  6 23  0    124  81944  92464  95516    0    0     0  1722  
> 620   790 61 39
> 0
>  7  0  0    124  72148  92472  95560    0    0     0  2358  
> 817  1199 62 38
> 0

It seems to be primarily cpu-bound (first column tells you how many processes are queued for cpu time), but it is not that bad.

> 
> >
> > Also, check your logs for timeouts.
> >
> > Do you have a caching DNS server on this machine?  If not 
> you should.
> >
> 
> There are a fair number of timeouts. 

Timeouts on what? RBL? Spamassassin?

>I do not have a caching 
> name server on
> this server, but I am going to load one on and see if that helps.

if redhat, install the packages called caching-nameserver

> I am
> running three RBL's on this system so their are a lot of DNS lookups
> happening.

hence the advantage of having a caching dns on the same machine.
> 
> Eric
>