Skip to main content

68 posts tagged with "Primary Site"

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

View All Tags

Primary Site v0.0.109

This release improves import reliability on index-in-place sites, fixes playback and export failures, speeds up file deletion, and stops trusting certificates from compromised eMudhra certificate authorities.

Changed
  • queryService.deployment.enableBetaStreamExecutor no longer has any effect. All playback uses the faster streaming mode enabled by default in v0.0.108, and you can no longer switch back to the previous mode. You can remove the setting from your values.yaml.
  • The time_to_first_byte_ms and stream_duration_ms query service metrics now measure the whole request, including authorization and request setup. Expect higher values after upgrading, and adjust any dashboards or alerts built on these metrics. See Query service metrics.
Fixed
  • On index-in-place sites, files that cannot be indexed because of a permanent problem, such as a corrupt or unsupported file, are now quarantined. Previously they were retried indefinitely. See Quarantined files.
  • On index-in-place sites, quarantined files can now be downloaded. Previously the download failed.
  • Deleting a file from an index-in-place bucket now removes its recording from Foxglove even when the site cannot list the bucket. Previously, AWS buckets without s3:ListBucket permission kept deleted recordings visible indefinitely.
  • Playback no longer fails for JSON messages that contain null values in an array.
  • Exporting a recording as a ROS 1 bag no longer produces a file the Foxglove CLI rejects.
  • When an inbox listener pod is stopped, such as during an upgrade or scale-down, it now finishes the imports it is processing before it exits.
Performance
  • Deleting recordings and cleaning up temporary files is faster on sites with many files.
Security
  • Primary site services no longer trust certificates issued by compromised eMudhra certificate authorities.

Docs: About | Install | Upgrade

Primary Site v0.0.108

This release improves search performance and enables the multi-threaded stream executor by default.

Performance
  • Added support for filter pushdown on recordings, devices, events, sessions, and device custom properties
  • Added support for skipping files when a search cannot return a hit for them
  • Improved how a search plans and orders its joins, including for recordings without a device
Changed
  • Enabled the multi-threaded stream executor by default. Set queryService.deployment.enableBetaStreamExecutor: false to fall back to the single-threaded reader
  • Improved support for streaming by recordingIds
Fixed
  • Fixed lookback dropping messages that share a channel's latest timestamp
  • Fixed streaming index-in-place recordings whose metadata or attachment sources are missing

Docs: About | Install | Upgrade

Primary Site v0.0.106

When resolving a recording's context, this release merges every "foxglove" MCAP metadata record in the file, rather than using only the last one. Later records override earlier keys, and an empty string value unsets a key. To keep the previous behavior, set the following in your values.yaml, on whichever component your indexing strategy deploys:

inboxListener:
  deployment:
    mergeFoxgloveMcapMetadata: false

This release also adds an optional beta stream executor, which can substantially improve streaming throughput on multi-core query service pods. It is off by default.

Added
  • Added a beta multi-threaded stream executor that reads a stream's MCAP chunks across parallel partitions, off by default behind queryService.deployment.enableBetaStreamExecutor
Changed
  • Merged all "foxglove" MCAP metadata records when resolving recording context, instead of using only the last record. Set mergeFoxgloveMcapMetadata: false on the indexer or inbox listener deployment to restore the previous last-record-only behavior
Fixed
  • Improved query service error messages when a stream fails mid-response
  • Fixed values.yaml validation, which was silently skipping its required-setting checks

Docs: About | Install | Upgrade

Primary Site v0.0.105

This release improves search performance and raises some default resource limits in the query server.

Changed
  • Defaulted the MAX_OPEN_FILES variable to 65535
  • Raised the default query service CPU limit from 1 to 4 and memory limit from 2Gi to 8Gi
Performance
  • Improved message scanning performance for search through multi-threaded scanning and other optimizations

Docs: About | Install | Upgrade

Primary Site v0.0.103

Added
  • Exposed the object store read timeout via the OBJECT_STORE_READ_TIMEOUT_SECONDS variable
Fixed
  • Fixed the AWS_COPY_PART_SIZE_BYTES variable, which was documented but not previously applied to S3 multipart copies; the default part size now returns to 100 MiB from 5 GiB

Docs: About | Install | Upgrade