Giovanni Mazzini@ESA
08/11/2000 09:58 AM
The IO timeout was not set.
Anyway I do not think it is a timeout problem.
In fact I made the following temptatives:
1) I set the timeout flag with a number of seconds of 10000 (over 2 hours and a
half) and the process (rsync server child) died long before
2) I commented the check_timeout() routine in the io.c file and the process
(rsync server child) died as usual.
The process does not die owing to SIGUSR1 or SIGUSR2 signals in fact I modified
the sigusr1_handler() and sigusr2_handler() in order to receive a message when
executed and I did not received any message.
I have not still included the patch you suggested me but I will do it as soon as
possible
Giovanni
David Bolen <[EMAIL PROTECTED]> on 2000/08/09 07:35:46 PM
Please respond to [EMAIL PROTECTED]
To: Multiple recipients of list RSYNC <[EMAIL PROTECTED]>
cc: (bcc: Giovanni Mazzini/esrin/ESA)
Subject: RE: Rsync server dies while receiving files
[EMAIL PROTECTED] [[EMAIL PROTECTED]] writes:
> In order to perform tests on large file systems I was transferring
> the directory '/usr' from one system to the other. The transfer has
> never completed: it usually stops after ~250 Mb (the complete '/usr'
> is about 450 Mb).
>
> I have tried to debug the problem and I have seen that the child
> process (the child of the child) receives a signal while performing
> the routine recv_file() and dies, the signal is probably the SIGUSR2
> signal sent to it from its parent process (see do_recv() routine in
> main.c)
One thought - are you running with an IO timeout (--timeout) set?
There was a problem mentioned on this list a while back where such a
timeout incorrectly gets utilized by both the parent and child
processes on the receiver - but because its only the child actually
receiving the data, the timeout becomes an overall process (as opposed
to an individual I/O) timeout for the parent which can then kill
things off if your overall transfer is long enough.
There was a small patch posted to this list (by Neil Schellenberger,
around the end of June I think) that eliminated the timeout in the
parent once the child had been created. It should be in the archives.
Yep - it's here: http://www.samba.org/listproc/rsync/July2000/0002.html
It's worked well for me.
-- David
/-----------------------------------------------------------------------\
\ David Bolen \ E-mail: [EMAIL PROTECTED] /
| FitLinxx, Inc. \ Phone: (203) 708-5192 |
/ 860 Canal Street, Stamford, CT 06902 \ Fax: (203) 316-5150 \
\-----------------------------------------------------------------------/
