smbd 2.2.0; Solaris; performance woes?

[David Collier-Brown]

David Lee wrote:
> On our two CPU Sun E-450, the old load was typically 1.3 to 2.6 ; the new
> load was typically always over 2, sometimes over 4, and sometimes much
> higher...

	As it's Solaris, can I get some sar data?
	use
		sar -o name.raw 10 30 && sar -A -f name.raw >name.txt
	to get a 5-minute sample during
		- a quiet period (call this "quiet.txt")
		- a normal busy period ("normal")
		- the mad rush as classes change and everyone
		logs in (or when a lab opens for the day)
		("login")
	sar -o doesn't load the system like vmstat does,
	it just writes raw data to a file. sar -f is compute-
	intensive, as it does many many printfs.
>
> For the moment I have had to revert to 2.0.7 (I suppose I should go to
> 2.0.9 if this is likely to prolong itself).

	I'd like the 2.0.7 sars, and I'll report if there is
	something tight with them, and then we can plan
	a test with 2.2: probably the week **after** the
	mid-term projects are due (;-))

> Anyone know of any problems on large installations?  I suspect it is
> scaling-related and that 2.2.x is more sensitive than 2.0.x .

	Quite likely...

--dave (I'm funded to do these during work hours) c-b
-- 
David Collier-Brown,           | Always do right. This will gratify 
Performance & Engineering Team | some people and astonish the rest.
Americas Customer Engineering  |                      -- Mark Twain
(905) 415-2849                 | davecb at canada.sun.com