[Samba] deadtime/keepalive not working as expected
Ed Siefker via samba
to unlock a file when I think it should have killed itself.
I have deadtime set to 15 and keepalive at default (300). I opened a file on
my server in Excel, then shut down the machine. Hours later someone
tells me the file is locked.
Shouldn't keepalive have detected that the machine was off in 5 minutes?
Shouldn't deadtime have disconnected the process and unlocked the file
15 minutes later? Do I misunderstand what is going on? Thanks
-Ed
--
To unsubscribe from this list go to the following URL and read the
instructions: https://lists.samba.org/mailman/options/samba
Jeremy Allison via samba
> My expectations are probably wrong, but I had to manually kill a process
> to unlock a file when I think it should have killed itself.
>
> I have deadtime set to 15 and keepalive at default (300). I opened a file on
> my server in Excel, then shut down the machine. Hours later someone
> tells me the file is locked.
>
> Shouldn't keepalive have detected that the machine was off in 5 minutes?
> Shouldn't deadtime have disconnected the process and unlocked the file
> 15 minutes later? Do I misunderstand what is going on? Thanks
Once the client is dead the TCP keepalive's should
detect the socket is down and thus shut off the servicing
smbd. Check this by doing a wireshark capture and a
debug level 10 on that smbd.
Volker Lendecke via samba
> On Tue, Nov 01, 2016 at 05:16:47PM -0500, Ed Siefker via samba wrote:
> > My expectations are probably wrong, but I had to manually kill a process
> > to unlock a file when I think it should have killed itself.
> >
> > I have deadtime set to 15 and keepalive at default (300). I opened a file on
> > my server in Excel, then shut down the machine. Hours later someone
> > tells me the file is locked.
> >
> > Shouldn't keepalive have detected that the machine was off in 5 minutes?
> > Shouldn't deadtime have disconnected the process and unlocked the file
> > 15 minutes later? Do I misunderstand what is going on? Thanks
>
> Once the client is dead the TCP keepalive's should
> detect the socket is down and thus shut off the servicing
> smbd. Check this by doing a wireshark capture and a
> debug level 10 on that smbd.
Is that still valid for SMB2? On Linux I always use socket options for
that these days:
socket options = TCP_NODELAY TCP_KEEPCNT=4 TCP_KEEPIDLE=240 TCP_KEEPINTVL=15
Volker
Jeremy Allison via samba
> On Tue, Nov 01, 2016 at 05:09:34PM -0700, Jeremy Allison via samba wrote:
> > On Tue, Nov 01, 2016 at 05:16:47PM -0500, Ed Siefker via samba wrote:
> > > My expectations are probably wrong, but I had to manually kill a process
> > > to unlock a file when I think it should have killed itself.
> > >
> > > I have deadtime set to 15 and keepalive at default (300). I opened a file on
> > > my server in Excel, then shut down the machine. Hours later someone
> > > tells me the file is locked.
> > >
> > > Shouldn't keepalive have detected that the machine was off in 5 minutes?
> > > Shouldn't deadtime have disconnected the process and unlocked the file
> > > 15 minutes later? Do I misunderstand what is going on? Thanks
> >
> > Once the client is dead the TCP keepalive's should
> > detect the socket is down and thus shut off the servicing
> > smbd. Check this by doing a wireshark capture and a
> > debug level 10 on that smbd.
>
> Is that still valid for SMB2? On Linux I always use socket options for
> that these days:
>
> socket options = TCP_NODELAY TCP_KEEPCNT=4 TCP_KEEPIDLE=240 TCP_KEEPINTVL=15
That's what I said - TCP keepalives should detect the client is down.
I think the deadtime handler is still installed on the event context
in SMB2, so that will fire if the client closed all resources went
away. If the server thinks the client still has a connection (share)
open the deadtime won't disconnect the client - but the TCP keepalives
should.