100% cpu utilization of Solaris x86
Richard Bollinger
rabollinger at home.com
Wed Nov 7 10:32:25 GMT 2001
Sorry to butt in, but we have seen similar pileups even with older versions of Samba, under similar
circumstances:
We run Samba 2.0.5a on Solaris 2.4/2.5 on Sun Sparc. Under heavy load, the client PC's, mostly
Win9x, send multiple connection attempts until one succeeds - but ALL of the attempts apparently get
queued up, spawning several extra smbd processes, most of which fail due broken client connections.
Key factors are the impatience of the clients, slowness in initial response from smbd coupled with a
overloaded server CPU.
The code deleting "dead locks" in locking_init() and locking_end(), has a CPU load effect of order
N^2, where N is the number of smbd processes running, assuming most processes hold at least a few
locks; so it's a good place to start.
I think the key fix will be in improving smbd process start up time so that the clients don't give
up prematurely.
Maybe someone can test the theory by putting a forced delay in the startup code?
Rich Bollinger
More information about the samba-technical
mailing list