Looks like we could not run this command: r...@leopard-6:/volumes# gcore -c /tmp/idmapd-core `pgrep idmapd` gcore: invalid content string '/tmp/idmapd-core' usage: gcore [ -pgF ] [ -o filename ] [ -c content ] pid ... r...@leopard-6:/volumes#
-----Original Message----- From: Nicolas Williams [mailto:[email protected]] Sent: Monday, March 02, 2009 1:39 PM To: John Keiffer Cc: [email protected] Subject: Re: [cifs-discuss] idmapd leaking? On Mon, Mar 02, 2009 at 10:50:40AM -0800, John Keiffer wrote: > I have an OpenSolaris kernel version #104 Build from Nexenta Systems, > and we are seeing a problem with idmapd using too much memory. It > looks like it could be leaking. The error message reported actually > says "idmapd/4 allocates 1457MB, which is more than configured > maximum". 1457MB is a lot. Too much. Could you: # svccfg -s idmap svc:/system/idmap> setenv UMEM_LOGGING transaction,contents,fail svc:/system/idmap> setenv LD_PRELOAD libumem.so.1 svc:/system/idmap> ^D # # svcadm refresh idmap # svcadm restart idmap And next time this happens # gcore -c /tmp/idmapd-core `pgrep idmapd` # echo ::findleaks|mdb /tmp/idmapd-core And post the output of mdb somewhere? There is one known leak: 6570029 possible memory leak around LDAP bind code If the leaks all involve function names with "sasl" in them then this is 6570029, and perhaps a lousy interaction with your environment is causing this to strike too often. Are you using domain or workgroup mode? Nico -- _______________________________________________ cifs-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/cifs-discuss
