Hello, i have a little problem here: Using Samba 2.2.2 with IBM ClearCase (2003) on a Sun Solaris 8 machine (2 cpus, 4G ram) (IBM does not support Samba 3.0 with Clearcase, maybe Samba 3.0 will fix the problem, but not able to test it yet). There are more than 100 smbd processes running in average. Load average of the maschine: under 1.0 in normal operation. But from time to time there are system locks, means: there are many smbd's running (over 50 and more) that produces an load average of over 50.0 ! These processes consume about 2% cpu time per smbd, resulting in: cpu state: 0% idle 10% user 90% kernel. Simply: the machine does not responded to any request log.smbd says: [2004/07/20 10:07:48, 0] locking/locking.c:delete_fn(252) locking : delete_fn. LOGIC ERROR ! Entry for pid 16008 and it no longer exists! [2004/07/20 10:07:48, 0] locking/locking.c:delete_fn(252) locking : delete_fn. LOGIC ERROR ! Entry for pid 23576 and it no longer exists! [2004/07/20 10:07:48, 0] locking/locking.c:delete_fn(252) locking : delete_fn. LOGIC ERROR ! Entry for pid 4160 and it no longer exists! [2004/07/20 10:07:48, 0] locking/locking.c:delete_fn(252) locking : delete_fn. LOGIC ERROR ! Entry for pid 22062 and it no longer exists! [2004/07/20 10:07:48, 0] locking/locking.c:delete_fn(252) locking : delete_fn. LOGIC ERROR ! Entry for pid 9311 and it no longer exists! [2004/07/20 10:07:48, 0] locking/locking.c:delete_fn(252) locking : delete_fn. LOGIC ERROR ! Entry for pid 9311 and it no longer exists! Seems that smbd's are terminated and restarted again?! But what can cause this behaviour? Are there any timeouts in smbd that can cause this? Maybe the problem arise, if someone want to access a clearcase element over a clearcase view that is shared over samba to the windows world. If the element's size is huge and maybe compressed, clearcase can take a while to extract the element from the vob and give a response back to samba. But i can not understand, why nearly *all* smbd's are influenced? Many thanks for any help Thomas Maier
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Maier, Thomas wrote: | Hello, | | i have a little problem here: | | Using Samba 2.2.2 with IBM ClearCase (2003) on a Sun Solaris | 8 machine (2 cpus, 4G ram) (IBM does not support Samba 3.0 | with Clearcase, maybe Samba 3.0 will fix the problem, but not able | to test it yet). There are more than 100 smbd processes running | in average. Load average of the maschine: under 1.0 in | normal operation. | | But from time to time there are system locks, means: there | are many smbd's running (over 50 and more) that produces | n load average of over 50.0 ! These processes consume about | 2% cpu time per smbd, resulting in: cpu state: 0% idle 10% | user 90% kernel. Thomas, The 2.2. branch is no linger under development or maintenance. However at the very least I would suggest upgrading to to Samba 2.2.9. I remember this code and when Jeremy was working on it. Changes are that 2.2.9 will be better for you. I don't remember the Solaris lock kernel bug was Solaris 7, 8, or 9 (or all three). But check with you Sun rep about the fcntl() kernel lock patch. Don't remember the patch # right now. Sorry. cheers, jerry - ---------------------------------------------------------------------- Hewlett-Packard ------------------------- http://www.hp.com SAMBA Team ---------------------- http://www.samba.org GnuPG Key ---- http://www.plainjoe.org/gpg_public.asc "...a hundred billion castaways looking for a home." ----------- Sting -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFA/Uo8IR7qMdg1EfYRAj/rAJ9aScH9DGqQUYBw/1OL/nGf78XvaACgjLrG fqetpaETctETm32G4oVHnqE=wab/ -----END PGP SIGNATURE-----
On Tue, Jul 20, 2004 at 10:22:02AM +0200, Maier, Thomas wrote:> Hello, > > i have a little problem here: > > Using Samba 2.2.2 with IBM ClearCase (2003) on a Sun Solaris 8 machine (2 > cpus, 4G ram) (IBM does not support Samba 3.0 with Clearcase, maybe Samba > 3.0 will fix the problem, but not able to test it yet). There are more than > 100 smbd processes running in average. Load average of the maschine: under > 1.0 in normal operation. > > But from time to time there are system locks, means: there are many smbd's > running (over 50 and more) that produces an load average of over 50.0 ! > These processes consume about 2% cpu time per smbd, resulting in: cpu > state: 0% idle 10% user 90% kernel.2.2.2 is old even for the 2.2.x series :-). Please upgrade either to 3.0.x or the last 2.2.x version (2.2.9 I believe). Jeremy.