Skip telemetry writes when no database is configured - #66
Conversation
With no `database` in the controller config, every metrics poll built an engine from the string "None", so SQLAlchemy raised "Could not parse SQLAlchemy URL from given URL string" once per instance every 5 seconds. Telemetry has nowhere to go without a database, so the writes -- and the metrics that only exist to feed them -- are now skipped, and the fact is stated once at startup.
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars. ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Advanced Run ID: Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
nickhuo
changed the title
|
This doesn't look like the OTLP exporter branch? The telemetry writes you are skipping won't be used in the future |
|
I just tagged Felipe somewhere about this but I can't find it lol, but if UI doesn't pull from that DB anymore you can just remove the code instead of adding a skip file from the telemetry_logger.py |
|
The agent metrics doesn't currently have a OTLP reciever though so just remove the runtime_information table logic |
What is SQLAlchemy used for here? The purpose of this PR is to keep the log from polling every five seconds and flooding other error msg |
…nfigured
Resolves conflicts from the Ventis->CanyonOS rename (CAN-255) landing on
main after this branch's CAN-283 fix:
- telemetry_logging.py: keep resolve_database_url()'s "empty means no
database" behavior, updated to read CANYONOS_DATABASE_URL (renamed from
VENTIS_DATABASE_URL on main).
- global_controller.py: adopt main's parallelized per-instance polling
(_poll_one_instance via ThreadPoolExecutor) and its new OTel row export,
layering in the database-configured guard around the legacy
send_runtime_information/send_agent_information calls so OTel export
keeps working even without a database while the legacy SQL writes stay
skipped. Dropped the now-redundant _write_telemetry helper and the
removed _agent_host_key in favor of instance_manager._routing_endpoint_for.
- project_id default: took main's `config.get("project_id")` (no default),
matching main's `_load_config` which now always mints one.
- Updated tests/test_global_controller_telemetry_skip.py for the
canyonos_core module paths, the CANYONOS_DATABASE_URL env var, and the
new _poll_one_instance entry point.
3 participants
If we assume the database URL is a prerequisite, we can ignore this pr. @Saaketh0 @iidsample
With no database configured, every metrics poll tried to build a SQLAlchemy engine from the string
"None". It failed the same way every 5 seconds — Could not parse SQLAlchemy URL from given URL string— burying the errors operators actually need to see.Skip telemetry when there’s no database and fix URL resolution