[net.unix] Advice requested regarding strange rdist error messages

pfeiffer@rsch.WISC.EDU (Phil Pfeiffer) (09/26/86)

The incident that prompted this message happened under Ultrix, but I think
I've also seen this problem up under 4.3BSD Unix.

Problem:  rdist occasionally quits with strange error messages. 
This output fragment, for example, is from an rdist job I ran today:

> updating host boursin.wisc.edu
> need to remove: work/Wonly
> need to update: /usr/pfeiffer/pfeiffer/mainl/travel
> need to update: /usr/pfeiffer/pfeiffer/mainl/Earthwatch
> need to install: /usr/pfeiffer/pfeiffer/mainl/Monly
> need to install: /usr/pfeiffer/pfeiffer/mainl/LOCK
> need to update: /usr/pfeiffer/pfeiffer/mainl/muzak/badmusic
> need to update: /usr/pfeiffer/pfeiffer/mainl/muzak/lyrics
> need to update: /usr/pfeiffer/pfeiffer/mainl/muzak/misc
> need to update: /usr/pfeiffer/pfeiffer/mainl/muzak/weasels
> rdist: /usr/pfeiffer/pfeiffer/mainl/muzak/rdist010562: No such file or directory
 
rdist exited with a "1" status.  A second run produced the same messages and
same exit status;  a third run finally succeeded.  

I was wondering whether the problem was write-access conflicts.  Even though 
I am running rdist with the "v" switch on  (don't output changes, just report
them), the program may still be trying to write-access files it touches. 
Furthermore, when this problem cropped up today, I was running another
simultaneous rdist that touched the same files, but from "the other side"
(for my application, I need to know what entries are only in A, what entries
are only in B, and what files have been changed between A and B).  In the
past, I think this rdist problem has cropped up when I was manually scanning
files at the same time an rdist job was touching them.

----

Any suggestions or comments would be most appreciated.
-- 
-- Phil Pfeiffer

...!{harvard,ihnp4,seismo,topaz}!uwvax!pfeiffer
(608) 263-7308

chris@umcp-cs.UUCP (Chris Torek) (10/07/86)

In article <2760@rsch.WISC.EDU> pfeiffer@rsch.WISC.EDU (Phil Pfeiffer) writes:
>Problem:  rdist occasionally quits with strange error messages. 
>This output fragment, for example, is from an rdist job I ran today:

[many messages deleted]
>>rdist: /usr/pfeiffer/pfeiffer/mainl/muzak/rdist010562: No such file
>>or directory
> 
>I was wondering whether the problem was write-access conflicts.

No, though you are close:

>... I think this rdist problem has cropped up when I was manually scanning
>files at the same time an rdist job was touching them.

The way rdist copies files about might seem rather strange at first,
but if you think about it you will see why it is done that way.
Rdist creates little temporary working files for everything it will
update before it actually goes about updating them.  If you are
running another rdist at the same time, in particular with the `-R'
flag, it will tend to spot these `anomalous' files and remove them.
The first rdist complains as soon as it tries to find the file
again, and your rdist command is told to exit(1).

It is best to run exactly one rdist at a time, and only from one
`central' machine, so that multiple rdists will not interfere with
each other.  If you are sneaky, you can use `special's to sequence
rdist runs, if you must have more than one `master'.
-- 
In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 1516)
UUCP:	seismo!umcp-cs!chris
CSNet:	chris@umcp-cs		ARPA:	chris@mimsy.umd.edu