New features:
- guardiand can now ask other guardians to re-observe specific messages using a signed
SendObservationRequestgossip message. For now, this is only supported for Solana, but will be extended to every chain we support. This supersedes the old hardcoded recovery.cfg mechanism and enables manual retry of messages that were failed to be observed during network outages (#776, #777).
Improvements:
-
guardiand will check whether a given message has already achieved consensus before attempting to send an observation. This avoids spamming the network with stale observations when a node is behind and catching up (#744).
-
Log level for logs sent to Google Cloud reduced to INFO. In an ironic twist, it turns out that they're really bad at doing full text search on large amounts of logs (yep, Google failing at search). We plan to move to a different solution based on the open source ClickHouse database, this will make the amount of logs more manageable in the meantime (#775).