Pulsegrid / Changelog
Platform releases ship continuously. Agent releases are versioned and listed here. Breaking changes get 90 days notice and a migration note.
Sampling was previously Team and above. It is now on the free tier too, because charging for the feature that reduces our own storage costs never made sense.
sampling.target_spans_per_secPods restarting within the same node lost their label set for up to 60 seconds, which broke dashboards filtered by label. The label cache is now rebuilt from the informer rather than lazily on first span.
Schedule an export of any table to S3, GCS or R2 on an hourly or daily cadence. Useful for long term retention beyond 13 months and for joining observability data against your warehouse.
Announced 90 days ago on 19 April. The old location at /etc/pulsegrid-agent/ is read with a warning until 1 January 2027 and then ignored. The installer migrates it automatically. Anyone managing config with Ansible or Puppet should update their path.
Fifth region live. Existing organisations cannot move region, so new ones only. If you need to relocate, talk to support and we will do an export and reimport with a short read-only window.
Statement timeout applied to the outer query only, so a heavy JOIN could run past it and hold a slot. Now enforced across the whole plan. Some previously slow queries will start returning a timeout error rather than eventually completing.
Go services are now instrumented without rebuilding, using uprobes on the standard library. Requires kernel 5.8 or newer. The old build-time approach still works and will be supported until at least 2028.