Hello,
I am looking for a suitable method for the following use-case: during test, I need to access the connection to the workstation that runs my LAVA worker and use it as a second namespace in order to be able to issue some commands on the the workstation (like playing an audio file). In other words I need to integrate the workstation as an auxiliary device needed for testing in the test job definition (like ssh <user>@<ws_ip_addr> and send some commands). Using LXC is not an option at this moment. At a first sight, the Multinode support would have the most of the needed features, but I did not find an example of such an usage, if there is any example you can point me to, I will be grateful.
Thank you in advance, Oana Ercus
On Tue, 16 Oct 2018 at 08:51, Oana Ercus oana.ercus@linaro.org wrote:
Hello,
I am looking for a suitable method for the following use-case: during test, I need to access the connection to the workstation that runs my LAVA worker and use it as a second namespace in order to be able to issue some commands on the the workstation (like playing an audio file). In other words I need to integrate the workstation as an auxiliary device needed for testing in the test job definition (like ssh <user>@<ws_ip_addr> and send some commands). Using LXC is not an option at this moment.
Use a service installed on the worker which can provide the files for download or streaming direct to the DUT using TCP/IP. Then run the software on the DUT to pull / wget the relevant files. This way the fact that this is actually the workstation / worker is completely irrelevant - any fileserver on the internet would do just as well - providing the relevant files were available there. lava-dispatcher recommends apache2, so you can use that to provide downloads.
The serial connection between the DUT and the worker is out of scope for this and will always be too slow.
At a first sight, the Multinode support would have the most of the needed features, but I did not find an example of such an usage, if there is any example you can point me to, I will be grateful.
The only reason for MultiNode would be if you need to synchronise a service on one node (using QEMU to launch a VM) with a client on another node. There are examples of such client:server MultiNode test jobs in the documentation.
The worker itself cannot and must not be thought of as a part of the test job - it is not. You can run a test job on the worker using LXC or QEMU. Do not attempt to make changes on the worker from a DUT. Hosting and serving data is fine, as is hosting a container or a virtual machine.
Thank you in advance, Oana Ercus _______________________________________________ Lava-users mailing list Lava-users@lists.linaro.org https://lists.linaro.org/mailman/listinfo/lava-users
-
Neil Williams -
Oana Ercus