instance start fails at the dpd_ensure step when one of the switch zones is unavailable

[askfongjojo]

The issue is somewhat unexpected as the two scrimlets are meant to provide redundancy. The error can be replicated when a scrimlet is brought down (e.g. by clean-slating and powering it down for expungement).

07:08:37.734Z INFO 390cc468-2358-4e95-a692-a771162c3054 (dropshot_external): request completed
    error_message_external = Internal Server Error
    error_message_internal = saga ACTION error at node "dpd_ensure": unable to find dendrite client for switch0
    file = /home/build/.cargo/registry/src/index.crates.io-6f17d22bba15001f/dropshot-0.12.0/src/server.rs:938
    latency_us = 756038
    local_addr = 172.30.2.6:443
    method = POST
    remote_addr = 172.20.17.42:61496
    req_id = 55f3ba9d-5f67-418c-b6d3-6759b3e36cfb
    response_code = 500
    uri = /v1/instances/vm1/start?project=test