High volume problem: stat: no such file or directory

[Lars Nordin]

On Tuesday 24 January 2006 01:38 pm, Bob Robison wrote:
> On Tue, 24 Jan 2006 12:32:35 -0600
> Bob Robison <bob.robison at swri.org> wrote:
> 
> [...]
> > After running a while, I have seen as many as 300+ rsync processes
> > remain active.  I have seen the client side process stay active for a
> > long time (7-10 minutes) while trying to transfer a few-K file.  I
> > 'shut off' the source of these transfers and eventually, the server
> > process catches up.  The last time I tried this I saw a 10 minute gap
> > in the timestamps showing up in the syslog file --- like something
> > was hung for that whole time, and then finally cleared up.
> 
> ... update: I found that it wasn't a 10 minute gap, but rather data out of 
order in the syslog output.  i.e. several messages that had been time-tagged 
at 18:07 showed up after 18:17 in the log file.  This may be another 
indication that there is an undesirable interaction with syslog.

FYI, by default syslog does synchronous writes to the file specified in 
syslog. So depending on how busy syslog is and how busy the disk(s) that 
contain that syslog file then you could run into some poor performance. You 
could prepend the log file in /etc/syslog.conf with "-" (like
"mail.*    -/var/log/mail.log") and see if performance improves.