I still need to do some more work to understand the series of backend failures that could trigger a 404 vs. 503 when more handoff nodes are available.
In my example I had more than ndata devices offline when I got the 404 and less than nparity handoffs available when I got the 503.
I think the client impact could be better advocated.
We want to be careful not to return an error when the client retrying the operation isn't going to make things any better.
I still need to do some more work to understand the series of backend failures that could trigger a 404 vs. 503 when more handoff nodes are available.
In my example I had more than ndata devices offline when I got the 404 and less than nparity handoffs available when I got the 503.
I think the client impact could be better advocated.
We want to be careful not to return an error when the client retrying the operation isn't going to make things any better.