Re: switchdev test

[Jiri Pirko]

Tue, Apr 18, 2017 at 12:18:36PM CEST, liali(a)redhat.com wrote:
> Hi all,
>
> When i running qos-005, it always print a error: RPC connection to machine machine1 timed
> out
>
> 2017-04-18 05:38:38        (machine1)        -   DEBUG: Executing: "echo
> "start" > /proc/net/pktgen/pgctrl"
> 2017-04-18 05:39:34        (machine1)        -   DEBUG: Killing command with  pid
> "51245"
> 2017-04-18 05:39:34       (localhost)        -    INFO: Result: FAIL
> 2017-04-18 05:39:34       (localhost)        -    INFO: Status message from slave:
> "RPC connection to machine machine1 timed out"
>
>
> Is this because the lnst-slave machine1 is too busy(because it is generating traffic) so
> the lnst-ctl lost connection to it?
> If i reduce the pktgen's thread count, this error will disappear. But then it
> can't generate enough traffic to make ETS to take effect.
> Anyone encountered this error ?  
Ccing LNST people.

switchdev test

[Ondrej Lichtner]

On Tue, Apr 18, 2017 at 12:20:28PM +0200, Jiri Pirko wrote:
>  Tue, Apr 18, 2017 at 12:18:36PM CEST, liali(a)redhat.com wrote:
>  >Hi all,
>  >
>  >When i running qos-005, it always print a error: RPC connection to machine machine1
> timed out
>  >
>  >2017-04-18 05:38:38        (machine1)        -   DEBUG: Executing: "echo
> "start" > /proc/net/pktgen/pgctrl"
>  >2017-04-18 05:39:34        (machine1)        -   DEBUG: Killing command with  pid
> "51245"
>  >2017-04-18 05:39:34       (localhost)        -    INFO: Result: FAIL
>  >2017-04-18 05:39:34       (localhost)        -    INFO: Status message from slave:
> "RPC connection to machine machine1 timed out"
>  >
>  >
>  >Is this because the lnst-slave machine1 is too busy(because it is generating traffic)
> so the lnst-ctl lost connection to it?
>  >If i reduce the pktgen's thread count, this error will disappear. But then it
> can't generate enough traffic to make ETS to take effect.
>  >Anyone encountered this error ? 
>  
>  Ccing LNST people.
>  _______________________________________________
>  LNST-developers mailing list -- lnst-developers(a)lists.fedorahosted.org
>  To unsubscribe send an email to lnst-developers-leave(a)lists.fedorahosted.org 
Hi,

never encountered this before.

One idea that comes to mind is whether or not the device used for the
Controller-Slave connection is affected by the packet generator.
On principle, this should be a completely isolated NIC, not used for
testing, so if you have some spare cpu cycles left, I don't see an
obvious reason why the connection would time out.

-Ondrej

switchdev test

[Jan Tluka]

Tue, Apr 18, 2017 at 12:20:28PM CEST, jiri(a)mellanox.com wrote:
> Tue, Apr 18, 2017 at 12:18:36PM CEST, liali(a)redhat.com wrote:
> >Hi all,
> >
> >When i running qos-005, it always print a error: RPC connection to machine machine1
> timed out
> >
> >2017-04-18 05:38:38        (machine1)        -   DEBUG: Executing: "echo
> "start" > /proc/net/pktgen/pgctrl"
> >2017-04-18 05:39:34        (machine1)        -   DEBUG: Killing command with  pid
> "51245"
> >2017-04-18 05:39:34       (localhost)        -    INFO: Result: FAIL
> >2017-04-18 05:39:34       (localhost)        -    INFO: Status message from slave:
> "RPC connection to machine machine1 timed out"
> >
> >
> >Is this because the lnst-slave machine1 is too busy(because it is generating traffic)
> so the lnst-ctl lost connection to it?
> >If i reduce the pktgen's thread count, this error will disappear. But then it
> can't generate enough traffic to make ETS to take effect.
> >Anyone encountered this error ? 
>
> Ccing LNST people. 
Hi, the error "RPC connection to machine machine1 timed out" could also
mean that the command took more time than a default timeout which is 60
seconds. Every command run through host.run() has this timeout set. If
it's exceeded, this message is printed.

You could try to extend the timeout of PktgenTx run by:

host.run(..., timeout=600)

which would override the timeout to 600 seconds.

-Jan

switchdev test

[Liang Li]

>  Hi, the error "RPC connection to machine machine1 timed
> out" could also
>  mean that the command took more time than a default timeout which is 60
>  seconds. Every command run through host.run() has this timeout set. If
>  it's exceeded, this message is printed. 
>  You could try to extend the timeout of PktgenTx run by: 
>  host.run(..., timeout=600) 
>  which would override the timeout to 600 seconds. 
yes, thanks everybody.