samba - May 2002 - smb_newconn error

Okay, newbie alert.

We're having a problem with a linux (2.2.20 smp) webserver (apache) that is
a samba 2.0.7-4) client. Oddly, both client and server are Linux. (I didn't
set them up)

There are 2 webservers that in theory are set up exactly the same. When a
certain one of them goes live, it generates an error:

	smb_newconn failed to allocate new packet buffer max xmit 65535

The server then soon fork bombs and the machine has to be rebooted to regain
control. If the server is *not* live, but Apache stressed, no problems
occur.

Is there a FAQ or searchable db I can look this up in? 

ciao!

leam

"Today's mighty oak is yesterdays nut that held it's ground."
--bumper
sticker

On Mon, 20 May 2002, Hall, Leam wrote:
> Okay, newbie alert.
> 
> We're having a problem with a linux (2.2.20 smp) webserver (apache)
that is
> a samba 2.0.7-4) client. Oddly, both client and server are Linux. (I
didn't
> set them up)
Switching to NFS may be a good idea.

> There are 2 webservers that in theory are set up exactly the same. When a
> certain one of them goes live, it generates an error:
> 
> 	smb_newconn failed to allocate new packet buffer max xmit 65535
smbfs needs to allocate a buffer that is as large as smbmount negotiated
that the maximum transfer size should be. smbmount uses the samba default
of 65535.

Try reducing this to one page using this setting in smb.conf:
	max xmit = 4096

(At least I think smbmount uses smb.conf for this value, if you get the 
 same value in the error message then smbmount needs some fixing)

> The server then soon fork bombs and the machine has to be rebooted to
regain
> control. If the server is *not* live, but Apache stressed, no problems
> occur.
Do you get any other kernel error messages?

What is the difference between "live" and not-live?

/Urban

> On Mon, 20 May 2002, Hall, Leam wrote:
> 
> > Okay, newbie alert.
> > 
> > We're having a problem with a linux (2.2.20 smp) webserver
(apache) that
is
> > a samba 2.0.7-4) client. Oddly, both client and server are Linux. (I
didn't
> > set them up)
> 
> Switching to NFS may be a good idea.
One of my first responses, too.
 
 
> > There are 2 webservers that in theory are set up exactly the same.
When
a
> > certain one of them goes live, it generates an error:
> > 
> > 	smb_newconn failed to allocate new packet buffer max xmit 65535
> 
> smbfs needs to allocate a buffer that is as large as smbmount negotiated
> that the maximum transfer size should be. smbmount uses the samba default
> of 65535.
> 
> Try reducing this to one page using this setting in smb.conf:
> 	max xmit = 4096
> 
> (At least I think smbmount uses smb.conf for this value, if you get the 
>  same value in the error message then smbmount needs some fixing)
Didn't see the max xmit setting in smb.conf. Do I need to add it?
 
> > The server then soon fork bombs and the machine has to be rebooted to
regain
> > control. If the server is *not* live, but Apache stressed, no problems
> > occur.
> 
> Do you get any other kernel error messages?
None that I could find.
 
> What is the difference between "live" and not-live?
"live" means open to outside http requests.
 
Unfortunately, this has been a lower priority problem for us, so the logs
(and my memory) are old and fading. I may need to rcreate the problem just
to catch the errors and be able to look at them fresh.

ciao!

leam