On Thu, 8 Aug 2002 at 9:14am, KEVIN ZEMBOWER wrote

> Is my understanding correct that the dump phase runs concurrently on
> all the hosts at the same time? Then, the hosts transfer the backup to

If you don't exceed your netusage.  Typically one dumper per host, too, 
btw.

> the tapehost. Is this concurrent, too? Finally, the tapehost spools to
> the tape, for which there's just one instance.

Dumps are done straight to the tape server over the network.  So, in 
theory, one dumper/client should be dumping straight to the server.

> Is there any way to determine where all the time is being spent? I
> remember using amstatus while a dump was under way, and seeing stats
> about what percentage of time was spent waiting for processes to
> complete. However, it looks like the amstatus is erased after the run
> completes.

Au contraire:

amstatus $CONFIG $LOGDIR/amdump.1

will give you the completed amstatus from last night.

> Estimate Time (hrs:min)    0:09
> Run Time (hrs:min)         1:12

Hmm is there a wrap in the code there?

The part you snipped should also show if there are any particular 
partitions that take a *long* time to dump.

-- 
Joshua Baker-LePain
Department of Biomedical Engineering
Duke University