[CentOS] Odd failure of smbd to start from init.d - CentOS 5.4

[Les Mikesell]

On 5/21/2010 9:44 AM, Whit Blauvelt wrote:
> On Fri, May 21, 2010 at 10:04:36AM -0400, Ross Walker wrote:
>
>> By any chance did someone add smbd to xinetd?
>>
>> If so then xinetd has the port open and the smbd process will not bind.
>
> Nope. Not sure that would explain why a slight difference in how it's
> invoked, through the same init.d script, makes the difference in whether it
> runs. That is:
>
> sh /etc/init.d/smb start (and "/usr/sbin/smbd -D")
>
> which always works from console, differs from
>
> /etc/init.d/smb start (and "service smb start" too)
>
> which doesn't ever work on this box, how? This is when smb starts with
> "#!/bin/sh" anyway. Only thing I can figure is that there may be a subtle
> difference in timing, a slowing down just enough to make the startup
> tolerant of hardware that's right on the margin. There's no significant
> difference (if any) in envars.

The only difference here 'should' be that explicitly running 'sh' will 
invoke your own shell aliases and search PATH to execute sh, where if 
you omit it you'll get the #!/bin/sh interpreter specified in the script 
itself.  Is there anything in your aliases or anything before /bin in 
PATH where the working shell might be found?

Or, perhaps this difference is coincidental and something is randomly 
killing smbd.  You might be able to see something if you comment out the 
nmbd startup in the script and
strace -f /etc/init.d/smb start
but it will only be useful if smbd dies and you can see a failing system 
call causing it.

-- 
   Les Mikesell
    lesmikesell at gmail.com