SMB2/SMB3 workloads for testing. (smbtorture?)

[Christof Schmitt]

On Mon, Jun 24, 2019 at 03:01:50PM -0500, Christopher O Cowan via samba-technical wrote:

> In recent years, we've set the MIN Protocol to SMB2, in production, but I
> had not been running that way in my lab.
> 
> Since, I have been using smbtorture and bench.nbench to simulate
> workloads, things broke when I made the change to Min Protocol.
> I'm getting "Failed to open connection -
> NT_STATUS_INVALID_NETWORK_RESPONSE".  I'm thinking either I don't know the
> magic combination of switches, or I'm going to have to change the code a
> bit.    Fully expecting right now, that a different loadfile is needed, as
> well.  (Unless I'm mistaken)

bench.nbench uses SMB1, not any of the newer protocol versions.

> I found two conference presentations in with the SambaXP stuff, but not
> sure that I fully grok things, yet.   My understanding is that nbench is
> just replaying packets previously captured using one of the VFS modules.

nbench basically reads a text file (loadfile) which describes the SMB
requests to issue on each connection. Then you specify how many
connections you want.

> I would appreciate any advice for generating a representative SMB2/SMB3
> workload, hopefully smbtorture.

I don't think there is a similar test for SMB2/SMB3 in smbtorture.

> Regards,
> Chris
> 
> PS: I'm also looking at dbench, as well, but I having a little trouble
> compiling it on a RHEL7 desktop.   Wading through that, right now, as well.

I managed to compile dbench with recent Samba versions. There are some
small changes required.

Christof