Hi Steve,

I want to use multinode api to support multi-client applying the share resource,and server distribute resource according to client's request.So when and which client request can't be determine.

I think if lava multinode api provide a addtional parameter(besides message id, it no need save to /tmp/lava_multi_node_cache.txt ),server block at same message id then according to message parameter to know which client make the requst and go on.

Steve McIntyre <steve.mcintyre@linaro.org> 于2019年5月14日周二下午8:52写道：

Hi Jack,

On Tue, May 14, 2019 at 06:30:10PM +0800, jack lu wrote:
>
>your example:
>role
> - server
>...
> steps:
> - lava-wait ipv4_1
> - lava-wait ipv4_2
>..
>The server will be blocked at lava-wait event till client lava-send ipv4_1 ..
>message.
>If only one client run "lava-send ipv4_2 ip=193" then server can't exit block
>status because it be blocked at "wait ipv4_1" event.

In that case, you need to make your tests more deterministic. Only
start a test with two clients defined if you know that both clients
are going to start. LAVA is designed to work with reliable,
deterministic tests.

--
Steve McIntyre steve.mcintyre@linaro.org
<http://www.linaro.org/> Linaro.org | Open source software for ARM SoCs