Hello Brian,

Well, things are still broken with 3.1.5. Handshaking between udp-receiver and 
usp-sender are not working properly.  I am sure you read Dann's  message, and I was 
wondering if you have any suggestions for me to try on my systems.

I tried changing flamethrower.conf with our working environment but it did not make a 
difference.  

Currently, I do not get the segmentation fault while trying to transfer data.  The 
problem now is that when the image is being downloaded,  the udp-receiver stops 
receiving data, starts a retry, but the udp-sender continues to attempt to send data 
(doesn't kill it's image.pid) until it timesout (5 to 10 minutes later), by this time, 
the 7 retries have passsed by and installation breaks.

Dann added changes to cvs, so if you can take a look at it when you get a chance and 
let me know what else I should try, that would be great.  

Another thing that I wanted to mentioned regarding twicks we made to 3.1.4 to get it 
to work with a big image (1.6GB or greater) was that we twick the initrd.img by 
increasing the number of inodes for the tmpfs staging area. We found out that with a 
big image, although we had enough space in the tmpfs staging area, we ran out of 
inodes for remaining files.  I don't think this is our problem here, but I thought it 
was worth mentioning it.    This change hasn't been done on my 3.1.5 imageserver, only 
on our current working environment with 3.1.4 SI

Thanks again for taking the time to help us with these issues and look forward to hear 
from you soon.

Carlos Ar�valo
hp
HPTC 
Phone: 603-884-3390
Fax:     603-884-5191
E-mail: [EMAIL PROTECTED]



-----Original Message-----
From: dann frazier [mailto:[EMAIL PROTECTED] 
Sent: Sunday, September 07, 2003 10:32 PM
To: Arevalo, Carlos
Cc: Brian Elliott Finley; [EMAIL PROTECTED]
Subject: Re: [Sisuite-devel] RE: SI 3.1.5 segmentation fault


On Sun, Sep 07, 2003 at 07:30:37PM -0400, Arevalo, Carlos wrote:
> what version of udpcast rpm were you using in your imageserver Dann?
> 
> any chance looking at the segmentation fault in our servers?

i spent quite a bit of time on your cluster this weekend - the segfault went away with 
my last fix - the issue you saw after that was due to an old gzip, and is fixed with a 
new one.  the binaries i used are installed on your devel imageserver (old ones are 
backed up w/ a dannf extension), and the code i changed is in cvs.

the binaries on your test server get to the point of installing the file tree, but it 
always seems to fail at the same point (~260 MB into it)- 
maybe you already know how to tweak flamethrower, or maybe brian has some ideas...

also note that the retry mechanism isn't working - the client exits immediately when 
it tries to receive a cast that is already in progress.  the client detects that a 
cast is already in progress, so it would be possible to check for this failure case by 
either looking at the udpcast log, or the exit status of udp-receiver (assuming this 
is a unique exit code - if not, well, seems like a reasonable patch.)

fyi, i tried both the old 200212xx version and the 20030611 versions in the initrd, 
and saw the same results.


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Sisuite-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/sisuite-devel

Reply via email to