Skipped content of type multipart/alternativeFrom ted.saylor at boeing.com Wed Jul 24 09:52:21 2002 From: ted.saylor at boeing.com (Saylor, Ted) Date: Mon Dec 1 12:39:03 2003 Subject: [Samba] High CPU on some smb processes. Message-ID: <B1763666D020C845B0B98005D43DE06A02659676@XCH-MES-99.sw.nos.boeing.com> Hello, First, as a first time poster to this list, I would like to thank the SAMBA team for this continuously improved product. If it was not for SAMBA, the files I serve might have been moved to NT - and I would be out of a job! I have three HP L2000 servers running as a service guard cluster, and there are 7 packages that run on the cluster. SAMBA (HP product B8725AA - A.01.08 aka 2.2.3a) runs locally on each box. Because of the moving packages, I have found it easier to register my package NetBios names with the central Boeing WINS system - therefore I am not running nmbd. I do not serve printing or need to participate in any other NT stuff that I believe nmbd is for. All I do is serve files - mainly lots of Unigraphics CAD files. This systems works, but there is a hitch. While at any given time I might have 50 smbd processes running, once or twice a day there will be a handful that really start clocking CPU. They are not runaways and do eventually stop clocking, but I would like to know what the client is doing to cause this. So my questions are: Is there a way to turn up the logging to see if the clients are doing "sweeps" (finds) in explorer? Can the client sweeps be stopped? What other kind of client request could peg CPU? Will 2.2.5 help? Is there a setting to lock out local directories from the clients (e.g. /usr, /etc, /var) other than dont decend? Is there a easy way to "fix" a environment where root dir=/ and wide links=yes? Ted Saylor UNIX Sys. Admin. M-F ~7AM-~3:30PM 480-891-7884 Desk 602-318-5565 Cell Phone/Pager ted.saylor@boeing.com