On Wed, 13 Jul 2005 18:00:15 +0200, Clive Menzies <[EMAIL PROTECTED]> wrote:
<snip> >> >> Anyone have some insight into this? Maybe you experienced something similar >> and >> can give me a few tips on how to resolve it? > >In /etc/samba/smb.conf there is a setting: ># Most people will find that this option gives better performance. ># See smb.conf(5) and /usr/share/doc/samba-doc/htmldocs/speed.html ># for details ># You may want to add the following on a Linux system: ># SO_RCVBUF=8192 SO_SNDBUF=8192 > socket options = TCP_NODELAY > >You could try uncommenting the penultimate line. > >Regards > >Clive > <snip> P.S. AH SWEET BUDDAH. 'TIS FIXED Well, I had typed a whole giant message about why, even after trying the above, it still didn't work, when I happened to stumble across the fix via google and the Gentoo forums (hey, I was desperate) I did appreciate the suggestion above (and in fact tried it) but I didn't notice any difference. However, deleting two registry keys in XP proved to be the ultimate solution. Apparently this "bug" was introduced in Service Pack 2 (SP2) Anyhow, I'm gonna try to be as verbose about everything as possible in the hopes that the next poor bastard who suffers from this problem will be able to find this post in the archives. First the important info: Nuclear : My Windows XP SP2 machine Sikozu: Debian GNU/Linux "Sarge" box. HP LaserJet 3550N: The Printer and of course CUPS & Samba. Two registry keys are at fault here. (Yay MS!) HKEY_CURRENT_USER\Printers\DevModePerUser and HKEY_CURRENT_USER\Printers\DevModes2 in DevModePerUser I have an entry of: \\Sikozu\Color_LaserJet_3550 <-- DELETED IT in DevModes2 I have two entries: \\NUCLEAR\\\Sikozu\Color_LaserJet_3550,LocalsplOnly <--- KEEP \\Sikozu\Color_LaserJet_3550 <--- DELETED IT Now, from what I can surmise is the fix: By deleting those two DELETED keys. I've told XP to spool to itself only. and then hand off the data to Samba _after_. This results in 100% CPU usage for about 10 seconds, and then *presto* I have my machine back. No more 5 minute freezes, etc. I'm sure someone else can probably shed more light the exact technical details, but long story short: It didn't work before, and now it does. --- PREVIOUS TEXT THAT I'M LEAVING FOR KEYWORDS --- Clive, Thanks for the tip. I gave that a whirl, but it didn't improve it any that I could notice. However, it did give me some things to think about. As near as I can tell (after running a couple of experiments) is that whatever is going on involves the XP box "talking" to the CUPS machine. For example, when printing a document from within Acrobat 7, the print dialog brings up a little print preview display so that one can see what the document will look like before it is actually printed and paper/ink are expended. Now, based upon my experiments, whenever the display is redrawn, the print driver itself communicates with the CUPS machine to redraw this little window. Or perhaps communicates is not the right word, maybe Query is more correct. Perhaps the driver is querying the print server to find out if a particular paper size or layout is supported, I don't know. But some sort of data transfer is taking place. (and a god awful amount of it too) As long that little display doesn't need to be redrawn (overlapping window) and no settings (orientation, page count, etc) are changed, the CPU & Bandwidth usage will eventually drop to zero and remain there. I can then proceed to press the OK button to initiate the print, and all goes as planned (and I get beautiful prints). But God forbid I change a setting (maybe I want 2 copies or I only want the first page printed, etc) and the whole cycle repeats. Network bandwidth jumps to 25mbps for 3-5 mins, CPU usage maxes out (which is due to the network traffic being generated I think). However, just as before, eventually the Print Preview window is redrawn and everything drops back to zero. I can click OK and get my printout. The really odd thing is that printing _directly_ to the printer doesn't result in the same kind of massive traffic. I might get 25mbps but only for a few seconds. So something is obviously jinky with Samba. Anyhow, if you or anyone else maybe has some more suggestions, I'd love to hear them, 'cause I'm at my wit's end. Thanks! -- Johann --