The fin has issues flashing right now because the TCP connection between the worker and core drops out while waiting for the fin to re-attach as a block device.
This is because the TCP connection must be considered Idle at this point, and times out.
I have managed to stop the failure from happening by sending the image in its entirety before initiating the fin flashing procedure - however this adds more time to the test.
I see several paths to fixing this:
rework the /flash endpoint of the worker so that the image to be flashed is sent in its entirety before the DUT is flashed, instead of using streams
Understand how to prevent the TCP connection from timing out. The worker does send keepalive packets at the moment - but they seem to be being ignored, and something is closing the connection.
Move the core back onto the testbot, as discussed.
jviotti
commented
6 years ago
The fin has issues flashing right now because the TCP connection between the worker and core drops out while waiting for the fin to re-attach as a block device.
This is because the TCP connection must be considered Idle at this point, and times out.
I have managed to stop the failure from happening by sending the image in its entirety before initiating the fin flashing procedure - however this adds more time to the test.
I see several paths to fixing this:
/flashendpoint of the worker so that the image to be flashed is sent in its entirety before the DUT is flashed, instead of using streams