[UI] When changing configuration of an Interface, one has to enter the config twice

[Jeff Lane ]

See the attached video.

If I have an interface that is configured, and I then change that configuration so it is put on a different fabric and subnet, when I click Save, instead of saving, MAAS changes the Interface to Unconfigured, forcing me to configure the interface a second time.

This is pretty easily and reliably reproducible on MAAS 2.6, I've had it happen on multiple systems when trying to move an interface from one Fabric to another.

This does not seem to occur when configuring one for the first time, or when simply changing the type of addressing (e.g. from DHCP to AutoAssign). It only seems to happen when I change one from one Fabric to Another, and choose a different subnet.

MAAS version: 2.6.0 (7802-g59416a869-0ubuntu1~18.04.1)

[Jeff Lane  (bladernr)]

Also, just to be sure I wasn't merely being impatient, I recreated this and let it sit for about 5 minutes in case this was just some sort of odd delay. But after 5 minutes the device I had modified was still in Unconfigured state, so I think we can rule out just a strange UI delay.

[Steve Rydz (steverydz)]

This seems to be an issue with the websocket. Saving these parameters seems to work only every other time. On the first attempt we don't receive the correct data from the websocket. On the second time the updated parameters data is received.

Here is an example of what is sent and received from the websocket when it doesn't work:
Sent: https://pastebin.canonical.com/p/9QsNd8h3v2/ (correct params are sent)
Received: https://pastebin.canonical.com/p/SfXwk3cSqJ/ (see that the links array is empty)

Here is an example of what is sent and received from the websocket when it does work:
Sent: https://pastebin.canonical.com/p/chwGgfPSBd/ (correct params are sent)
Received: https://pastebin.canonical.com/p/mZmCkVN2gf/ (see that the links array has the correct data)

[Jeff Lane  (bladernr)]

Is there any ETA on getting this resolved in MAAS?

[Huw Wilkins (huwshimi)]

This appears to still be happening. We're sending the same websocket message both times, so maybe there's an issue on the API side.

[Caleb Ellis (caleb-ellis)]

Still broken as of 2022-05-11