Agreed.
On Thu, 20 Jan 2000, Gustav Schaffter wrote:
> Charles,
>
> I don't know how logrotate does, but this scheme would lose data if (on
> a busy system) the process is actively writing to the log file in the
> moment of the 'rotate' action.
> >From the moment you have made a copy until you truncate the contents of
> the file, a number of lines (and/or part of a line (shiver)) may have
> been written to the log file. They would all get lost.
>
> I repeat: I don't know how logrotate does it.
>
> But I hope it's something more secure.
>
> Best regards
> Gustav
>
> Charles Galpin wrote:
> >
> > I've been reading this thread with much interest. It was my understanding
> > that you can safely do
> >
> > cat /dev/null > file
> >
> > and not destroy the inode, so why can't logrotate (or anything
> > else) simply do
> >
> > cp file file.1
> > cat /dev/null > file
> >
> > as a rotation scheme without restarting or signaling the process that is
> > writing to the file?
> >
> > thanks
> > charles
--
To unsubscribe: mail [EMAIL PROTECTED] with "unsubscribe"
as the Subject.
