strange ps behaviour

[Michael Lednev]

Oliver Fromme пишет:
> Michael Lednev wrote:
>  > Oliver Fromme ?????:
>  > > Michael Lednev wrote:
>  > > > Oliver Fromme ?????:
>  > > > > Michael Lednev wrote:
>  > > > > > # pgrep radiusd
>  > > > > > 1105
>  > > > > > 33738
>  > > > > > # ps ax | grep radiusd
>  > > > > >  1105  ??  Ss     2:35,76 /usr/local/sbin/radiusd
>  > > > > > # ps 33738
>  > > > > >   PID  TT  STAT      TIME COMMAND
>  > > > > 
>  > > > > It seems that the second matching process exited before
>  > > > > the ps command was executed.
>  > > > 
>  > > > It's repeatable.
>  > > 
>  > > That means that the radiusd process kept forking short-
>  > > lived child processes, for whatever reason.
>  > 
>  > keeping constant PID for children?
>
> OK, you didn't mention that it is the same PID every time.
> In that case my first suspicion would be a bug in pgrep.
> If it happens again, I suggest you use pgrep -lf.  Maybe
> the output gives a hint.
>   

It happened again.

# pgrep -lf radiusd
74847 /usr/local/sbin/radiusd
93248 radiusd

There is no /proc/93248 directory, so I assume there's surely no such 
process. Am I right?

> (Also note that ps cuts after 80 columns.  Sometimes the
> information you're looking for is after column 80, so I
> recommend to always use -ww, especially when the output
> is used for matching in scripts.)

I use pgrep in my script. If I need ps I always use ww, but thanks for 
suggestion anyway.