MGS/sp-sim: Introduce location mapping #933
Conversation
| /// List of human-readable location names; the actual strings don't matter, | ||
| /// but they're used in log messages and to sync with the refined locations | ||
| /// contained in `determination`. For "rack v1" we could use something like | ||
| /// "sidecar-a"/"sidecar-b". |
There was a problem hiding this comment.
let's ask @rmustacc what the printing on the rack is going to say and match that?
There was a problem hiding this comment.
Robert pointed me to RFD 200 § 7.2, which refers to them as SWITCH 0/SWITCH 1 (if space permits) or SW0/SW1 (shorthand). I'm slightly inclined to stick with lowercase for log strings/config files, so maybe switch0/switch1?
| port_config.location.remove(&location).ok_or_else(|| { | ||
| StartupError::InvalidConfig { | ||
| reason: format!( | ||
| "port {} has no entry for determined location {:?}", | ||
| port.0, location, | ||
| ), | ||
| } | ||
| })?; |
There was a problem hiding this comment.
might we want to see all such errors rather that just the first? Or perhaps log out all mismatches and return an error saying that the config was bad?
There was a problem hiding this comment.
I considered doing some kind of full-config-file-validation up front; that seems better than failing here after we've already talked to some SPs. I'll do that and log all errors.
There was a problem hiding this comment.
Added config validation in 674ffc2; it's a little long/boilerplatey but I think that's probably fine in exchange for getting a complete list of errors.
| 0 => determination.sp_port_0, | ||
| 1 => determination.sp_port_1, |
There was a problem hiding this comment.
do I recall correctly that the KSZ numbers these 1 and 2? If that's the case we might want to stick with that to prevent confusion in the future.
There was a problem hiding this comment.
Fixed in 7782fbc
jgallagher
force-pushed
the
mgs-sp-discovery-prep1
branch
from
3aa03c0 to
a089bb4
Compare
This runs the SP socket address and RFD 250 location discovery process.
jgallagher
force-pushed
the
mgs-sp-discovery-prep2
branch
from
b098ac2 to
674ffc2
Compare
This PR builds on #932 and is the second of three building toward a preliminary implementation of RFD 250-style SP and location discovery.
There are two relevant changes in this PR (plus some refactoring of error types):
Ping/Pongmessages are replaced withDiscover/DiscoverResponse, the latter of which includes the port on which the SP received theDiscover.LocationMaptype is added (but not currently used - that will be in the next PR). This includes the tokio task responsible for sendingDiscoverpackets on all switch ports, collecting the responses, and using them to derive the current location based on the configuration. The configuration includes a very simple description that lets us implement RFD 250 section 4.4: it lists specific ports we want to ping, and what the responses on those ports mean in terms of which sidecar we're running on.