Open mhofman opened 4 hours ago
Yeah, that sounds reasonable to me. If it's easy, I'd add owner
to the bridge-inbound records, that's one of the things I've needed to extract from the delivery to add to details
fields (like identifying which oracle is submitting a PushPrice).
I think timer.poll()
returns a boolean (wokeAnything
in device-timer.js
), not an integer, but still, including it in the slog event sounds useful.
Tell me more about "messages/acks info on cosmic-swingset-deliver-inbound
" ? I don't think we have an event of that type. Are you thinking about whether a given bridge-inbound
succeeded or failed? Or are you thinking about the sizes of the pending-sends and the other hangover-prevention data structures?
I'd add
owner
to the bridge-inbound records, that's one of the things I've needed to extract from the delivery to add todetails
fields (like identifying which oracle is submitting a PushPrice).
I believe all that is included in the body
. See doBridgeInbound
I think
timer.poll()
returns a boolean (wokeAnything
indevice-timer.js
), not an integer, but still, including it in the slog event sounds useful.
Yeah whatever the return value is. The name is addedToQueue
in cosmic-swingset, which is not clear what type it was.
Tell me more about "messages/acks info on
cosmic-swingset-deliver-inbound
" ? I don't think we have an event of that type.
I'm talking about the deliverInvound
of the mailbox, which arguably we don't really see on mainnet but that is technically enabled at that level (but are not provisioned so splats soon after I think).
Also I just remembered that any new events need to be added to the otel slog sender to avoid the problem of #8272/#9569 (which we likely should fix while we're at it).
What is the Problem Being Solved?
Timer polls and install bundles generate swingset runs (possibly empty), but unlike "bridge" and "deliver" (mailbox) triggers, we do not record a trigger event for these. This forces any contextualized slog processor (e.g. #10269 or #8680) to peek ahead at the first delivery to classify the type of run. We should also consider including more details in the existing trigger to enhance the classification.
Description of the Design
cosmic-swingset-timer-poll
slog event. Include theblockTime
of the poll, and the number of wakes added to the queue.cosmic-swingset-install-bundle
slog event. Include theendoZipBase64Sha512
of the bundler, and anyerror
.cosmic-swingset-bridge-inbound
slog events.cosmic-swingset-deliver-inbound
slog eventsTo avoid future regressions with missing triggers for swingset runs, we should consider plumbing this into
runSwingset()
somehow.Security Considerations
None
Scaling Considerations
A little extra/duplicated data, but very little in comparison of the rest of the slog size.
Test Plan
Manual is good enough
Upgrade Considerations
Requires a change of the chain software, but it does not affect consensus.