deadtime has no effect, max smbd procs has no effect

[Richard Sharpe]

On Sun, Jan 12, 2014 at 2:15 PM, Kartik Vashishta <kartik.unix at gmail.com> wrote:
> Ok. Thanks a lot. We can consider upgrading samba to a major release in the
> 3.x tree ( not 4 yet). WE do not have the debuginfo RPM installed for samba,
> how do I do that?

The debuginfo RPMs might already be installed. If not, they should be
available. I seem to recall using on a CentOS 6.x system.


> On Sun, Jan 12, 2014 at 1:47 PM, Richard Sharpe
> <realrichardsharpe at gmail.com> wrote:
>>
>> On Sun, Jan 12, 2014 at 10:25 AM, Kartik Vashishta
>> <kartik.unix at gmail.com> wrote:
>> > smbstatus shows one TCP connection, or at most 2.
>>
>> JSYK, samba 3.5.10 is very old and does not support SMB2.
>>
>> So, how are you detecting these smbds?
>>
>> Have you considered gdb -p <pid-of-one-of-the-errant-smbds>
>>
>> This should allow you to attach and if you have the debuginfo RPM
>> installed for Samba you should be able to see where it is stuck/hung
>> or what it is doing.
>>
>>
>> > On Sun, Jan 12, 2014 at 12:19 PM, Richard Sharpe
>> > <realrichardsharpe at gmail.com> wrote:
>> >>
>> >> On Sun, Jan 12, 2014 at 9:56 AM, Kartik Vashishta
>> >> <kartik.unix at gmail.com>
>> >> wrote:
>> >> > We have what we believe to be a design issue with our application
>> >> > configuration, leading to:
>> >> > deadtime = 0; having no effect - perhaps number of open files is not
>> >> > zero;
>> >> > lsof indicates this
>> >> > max smbd processess = <some value>; has no  effect, this preplexes us
>> >> >
>> >> > smbstatus -p
>> >> > Samba  version 3.5.10-125.el6
>> >> >
>> >> > Our normal smbd processes should be less than 10, however they climb
>> >> > well
>> >> > beyond the limit in max smbd processes in smb.conf
>> >> >
>> >> > We are running CentOS 6.3
>> >> >
>> >> > Any help would be appreciated; any workaround?
>> >>
>> >> Does smbstatus also indicate that those smbds have active tcp
>> >> connections. That is, you will have to look at the state of their
>> >> connections.
>> >>
>> >> Also, those smbds could be stuck in the file system, and there is a
>> >> bug in that version where it does not notice on write that the
>> >> connection is gone, and if there is a lot of data in the socket
>> >> buffers, the smbds can hang around for a while.
>>
>> --
>> Regards,
>> Richard Sharpe
>> (何以解憂？唯有杜康。--曹操)
>
>



-- 
Regards,
Richard Sharpe
(何以解憂？唯有杜康。--曹操)