Skip to main content

33 posts tagged with "Primary Site"

Release notes for self-hosted Foxglove Primary Site Helm chart.

View All Tags

Primary Site v0.0.49

This release contains minor updates and fixes to the inbox listener and garbage collector.

New & Improved

  • Added: If the LOG_REPEATED_IMPORT_FAILURES environment variable is set to true, the inbox listener will log an error-level message whenever it has to quarantine an input file after exhausting all retries. This can help Primary Site operators detect issues with their deployment.

  • Changed: For Primary Sites using Azure storage buckets with hierarchical namespaces enabled, the garbage-collector service will log an info-level message when it attempts to delete a non-empty directory, but it will skip over it.

    Fixes

  • Fixed: If the inbox listener attempts to process an MCAP file with an invalid file offset, it will quarantine the file immediately rather than retrying three times.

Docs: About | Install | Upgrade

Primary Site v0.0.48

This release adds configuration options for inbox listener autoscaling.

New & Improved
  • Added: Helm chart values to tune pollingInterval, successJobsHistoryLimit and failedJobsHistoryLimit for autoscaling

Docs: About | Install | Upgrade

Primary Site v0.0.46

This release contains support for downloading and retrying quarantined files.

New & Improved

  • Added: support for downloading quarantined pending imports. This feature will be enabled through the UI in the near future.

  • Note: For this to work, deployers will need to double-check permissions on their inbox bucket to make sure that their stream service has access to read from them.

  • Added: support for retrying quarantined pending imports. This feature will be enabled through the UI in the near future.

    Fixes

  • Fixed: if the garbage collector cannot delete a temp file, it will log an error instead of aborting.

  • Fixed: the request ID tag in inbox listener logs now matches the request ID in the foxglove API for pending imports.

  • Fixed: various corrupt BAG and MCAP errors will now result in the file being quarantined immediately rather than retrying 3 times.

Docs: About | Install | Upgrade

Primary Site v0.0.45

This release contains a fix to the inbox listener to support different storage providers for inbox & lake buckets.

Fixes
  • Fixed an issue where the inbox listener uses the configured inbox storage provider for metadata instead of lake

Docs: About | Install | Upgrade

Primary Site v0.0.44

This release contains a change to the inbox listener to write lake files with a new object path format.

Changes
  • The inbox listener now writes files to the lake bucket with a format starting with v1/.

Docs: About | Install | Upgrade

Primary Site v0.0.43

This release contains a fix to the inbox listener for metadata writes to the lake.

Fixes
  • Fixed an issue where the inbox listener would ignore failures to write metadata to the lake in some cases.

Docs: About | Install | Upgrade

Primary Site v0.0.42

This release contains a fix to the inbox listener for self-managed sites when importing recordings with no device ID and no key.

Fixes
  • Fixed an issue in the inbox listener where recordings with no device ID and no specified key were incorrectly de-duplicated.

Docs: About | Install | Upgrade

Primary Site v0.0.41

🗂️ Sorting messages by log time​

The inbox listener will now sort messages in files by their log time while importing. This means that if messages are recorded out of order because of multithreading, clock skew, or other reasons, they can still stream back in time-stamp order. See the inbox listener configuration documentation for configuration options.

New & Improved
  • Updated the inbox listener so if a file is provided with an invalid device name or ID, the import will fail immediately rather than being retried repeatedly
  • Added support for importing and streaming MCAP files containing channels with schema ID 0 (no schema)
Fixes
  • Fixed an issue where the stream server would exit uncleanly and log an error then the SHUTDOWN_TIMEOUT was exceeded while waiting for open requests to complete after being sent SIGTERM

Docs: About | Install | Upgrade