[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: 20000404: pbuf_flush (cont.)

On Tue, 4 Apr 2000, Unidata Support wrote:

> 
> ------- Forwarded Message
> 
> >To: address@hidden
> >From: "Neil R. Smith" <address@hidden>
> >Subject: 20000404: pbuf_flush
> >Organization: Dept. Atmospheric Science, TAMU
> >Keywords: 200004041712.LAA05951 LDM IDD ACTION log write failure
> 
> Robb Kambic wrote:
> > 
> > On Thu, 30 Mar 2000, Gregory Grosshans wrote:
> > 
> > > Can someone explain what a 'pbuf_flash 15' entry in the ldmd.log file
> > > means?
> > >
> > > This LDM ingests the Gateway channel from NOAAPORT (as FSL5) and PPS,
> > > DDS   (note request
> > > patter below).  The ldm queue is 750 MB.
> > 
> > Greg,
> > 
> > The pbuf_flush 15 means that the ACTION ie ( FILE, STDIO, PIPE, etc )
> > could not write to the destinations because the resource was not available
> > at that time or it was blocked.  The log entry ususally occurs when disk
> > activity is at it's peak for some reason.  It could be the LDM doing many
> > disk writes or another process tying up the disk.
> > 
> > time elapsed 5.690810
> > 
> > means that it took 5.6 sec to complete the activity,
> > 
> > If this occurs a bunch of time, one could start to loose data.  Then it's
> > time to determine the reason for the pbuf_flush entries.
> > 
> > Robb...
> 
> Well, we're getting quite a few of these.  
> 
> Sample hourly cumulative counts:
> 
> Apr 04 00     72
> Apr 04 01     39
> Apr 04 02     85
> Apr 04 03     104
> Apr 04 04     11
> Apr 04 05     27
> Apr 04 06     24
> Apr 04 07     35
> Apr 04 08     37
> Apr 04 09     56
> Apr 04 10     50
> Apr 04 11     44
> Apr 04 12     85
> Apr 04 13     76
> Apr 04 14     106
> Apr 04 15     143
> Apr 04 16     122
> 
> What's the next step?
> Thanks, -Neil


Neil,

At this point I would do nothing unless you are loosing data. If that's
the case then you might want to look into faster disk drives or controls
on the hardware side.  The other solution is to decrease the number of
writes by using less decoders, etc. One can off load the writes to another
LDM on a different machine also, just feed the second machine from the
first.  

Robb...



> -- 
> Neil R. Smith, Comp. Sys. Mngr.               address@hidden
> Dept. Atmospheric Sci., Texas A&M Univ.       979/845-6272 FAX:979/862-4466
> 
> 
> ------- End of Forwarded Message
> 

===============================================================================
Robb Kambic                                Unidata Program Center
Software Engineer III                      Univ. Corp for Atmospheric Research
address@hidden             WWW: http://www.unidata.ucar.edu/
===============================================================================