Re: Why systemtap loses events and what can be done about it?

[Brad Peters <bpeters at linux dot vnet dot ibm dot com>]

Frank Ch. Eigler wrote:
> Hi -
>
>   
> > <zzh@ncic.ac.cn> wrote:
> >     
>
>   
> > [...]  To make my question clear, I have to introduce my work in
> > more detail.  I am trying to use systemtap to analyze the
> > communication pattern of internet service. One basic issue is to
> > correlate every pair of tcp_sendmsg and tcp_recvmsg.
> >     
>
> (For what it's worth, there may not be a one-to-one correlation,
> considering signals, packet loss, buffered delivery, ...)
>
>
>   
> > Usually I will run a workload to emulate 100-200 clients explore the
> > web site simultaneously. In the workload of 200 clients, staprun
> > often displays the warning messages about skipped probes(Why?).
> >     
My guess would be this is caused by excess time spent in the probe 
handler; as interrupts are disabled whenever a probe is run, only one 
can run at any given time on a per cpu basis.  The only solution, AFIK, 
is to optimize/minimize your probe code.  This does not eliminate the 
problem, however.  Given enough probe hits, you will still see contention.

Best,

Brad Peters