What carverauto/serviceradar shipped
Written by FoxPlug from the repository's own releases; not affiliated with Serviceradar. Paste your own repository in the box below and FoxPlug writes the same weekly update for your product. This page is an automatic summary of the releases, pull requests and commits in github.com/carverauto/serviceradar; Serviceradar did not write it and does not use or endorse FoxPlug. Every line links to the change it describes.
Get a weekly update like this for your product, free
Week of November 2, 2026
What shipped
- V1.5.1 cut, reducing StarRocks load, isolating interactive queries from background work, and fixing ways the durable edge pipeline could lose data silently. Release
- StarRocks background warehouse load was cut and interactive queries isolated so the warehouse no longer runs out of memory under its own maintenance. Pull request #6390
- Edge records and sweep batches are no longer silently dropped on write failures. Pull request #6391
- The unreached legacy core ingest tree was retired. Pull request #6389
- The flow details modal now reads StarRocks process attribution instead of showing no match. Pull request #6408
- Helm install --wait no longer deadlocks on a fresh release, and Helm ownership is stamped on CNPG secrets. Pull request #6407
- Agent release import and publish now run in the background with a spinner instead of hitting the LiveView push timeout. Pull request #6409
- GitHub release asset downloads retry the other URL after a timeout, and redundant native add-on sync polls are skipped. Pull request #6415
- Settings -> Cluster now shows a per-contract edge-record pipeline so durable data families are visible. Pull request #6393
- StarRocks schema changes now wait out FINISHED_REWRITING before the next migration. Pull request #6375
Why it matters
This week centers on v1.5.1, which lowers the load ServiceRadar puts on StarRocks and keeps interactive queries apart from background work. It also closes several silent data-loss paths in the durable edge pipeline, including dropped edge records and sweep batches on write failures, and retires the unreached legacy ingest tree. Camera relay operations gained reconciliation so lost replies for opens and stops can be investigated instead of returning misleading errors.
Changelog entry
- Release v1.5.1 with reduced StarRocks load and resource-group isolation. Release
- Cut background StarRocks load and isolate interactive queries. Pull request #6390
- Stop dropping edge records and sweep batches on write failures. Pull request #6391
- Retire the unreached legacy core ingest tree. Pull request #6389
- Pin StarRocks sessions late, poll the migration lock, and row-bound bucket rewrites. Pull request #6401
- Store OTel span, BMP and flow-attribution text whole and null invalid identifiers in StarRocks. Pull request #6397
- Wait out FINISHED_REWRITING before the next StarRocks schema change. Pull request #6375
- Pin StarRocks 4.1 installs to 4.1.7 and document upgrade hops. Pull request #6394
- Unblock helm install --wait and stamp Helm ownership on CNPG secrets. Pull request #6407
- Run agent release import and publish in the background with a spinner. Pull request #6409
- Read StarRocks process attribution in the flow details modal. Pull request #6408
- Expose a camera relay stop reconciliation API. Pull request #6411
V1.5.1 is out: less StarRocks load, interactive queries kept apart from background work, and no more silently dropped edge records or sweep batches. Legacy core ingest tree retired. Camera relay opens and stops can now be reconciled after a lost reply.
ServiceRadar v1.5.1 is out. It cuts the load the project puts on StarRocks and keeps interactive queries separate from background maintenance, so the warehouse no longer runs out of memory under its own work. It also stops two ways the durable edge pipeline could lose data silently, including dropped edge records and sweep batches on write failures. The unreached legacy core ingest tree is retired, Helm install --wait no longer deadlocks on fresh releases, and camera relay opens and stops can be reconciled after a lost reply. StarRocks installs are pinned to 4.1.7.
Week of October 12, 2026
What shipped
- Flows are now stored only in StarRocks, dropping the legacy dual-write to the CNPG hypertable. Pull request #5736
- NVD gzip shards are read through end of input, so the nist-nvd2 advisory feed stops failing every cycle. Pull request #5723
- A full nist-nvd2 advisory feed load now finishes within preflight instead of timing out on the 580 MB corpus. Pull request #5746
- Advisory feeds load in per-chunk transactions, so no run is one huge transaction and unchanged corpora write nothing. Pull request #5761
- The flow collector now applies sampling rates that IPFIX and NetFlow v9 exporters declare in options records. Pull request #5707
- IPFIX exporters that declare whole-flow selection keep a sampling rate of 1 instead of being scaled. Pull request #5731
- Camera WebRTC signaling is routed to the core-elx node that hosts the relay. Pull request #5706
- Live camera video is gated behind a new devices.cameras.view permission. Pull request #5701
- OTLP log bodies longer than 4096 characters are kept as text instead of being exported as integer arrays. Pull request #5734
- The agent gateway stops re-pushing a config version it already sent on the same stream. Pull request #5728
Why it matters
This week fixes several long-running failures in the advisory and inventory pipeline, with faster NVD feed loading and smaller stored advisory payloads. Flow handling becomes StarRocks-only and now honors exporter-declared sampling rates, which matters to anyone reading NetFlow and IPFIX volume. Camera relay behavior and permissions are also tightened.
Changelog entry
- Store flows only in StarRocks and remove the legacy CNPG dual-write. Pull request #5736
- Read NVD gzip shards through input EOF so the nist-nvd2 feed stops failing each cycle. Pull request #5723
- Make a full nist-nvd2 advisory feed load finish within preflight. Pull request #5746
- Load advisory feeds in per-chunk transactions with in-memory presence. Pull request #5761
- Store each NVD configuration expression once per advisory to cut coordinate storage size. Pull request #5759
- Apply IPFIX and NetFlow v9 options sampling rates in the flow collector. Pull request #5707
- Keep whole-flow IPFIX selection at rate 1. Pull request #5731
- Route camera WebRTC signaling to the relay host node. Pull request #5706
- Gate live camera video on the devices.cameras.view permission. Pull request #5701
- Keep long OTLP log bodies as text rather than integer arrays. Pull request #5734
- Skip re-pushing an outstanding config version on the control stream. Pull request #5728
- Align package set hashes without ecosystem so unchanged uploads stop forcing full re-uploads. Pull request #5733
This week: flows move to StarRocks only, the NVD advisory feed loads fast enough to finish preflight, IPFIX options sampling rates are applied, camera WebRTC signaling goes to the relay host, and live video is behind a new permission.
This week the project made several pipeline fixes. Flows are now stored only in StarRocks instead of being dual-written to the legacy table, and the flow collector applies sampling rates that IPFIX and NetFlow v9 exporters declare in options records. The NVD advisory feed load now finishes within preflight, reads gzip shards through end of input, and loads in per-chunk transactions. On the camera side, WebRTC signaling routes to the node hosting the relay, and live video is gated behind a new devices.cameras.view permission. OTLP log bodies over 4096 characters are kept as text.
Week of September 14, 2026
What shipped
- StarRocks warehouse now serves NetFlow and telemetry data as an opt-in alternative to CNPG, with per-dataset configurable TTLs. Pull request #4485
- Network Flows page reduced from 14 serial loaders to one query, fixing slow page loads despite fast warehouse responses. Pull request #4499
- Topology graph replaced Apache AGE with Dgraph as the intended store, keeping AGE as a shadow during rollout. Pull request #4490
- Dashboard NetFlow map now draws every arc instead of limiting to 26, showing all 4,799 conversations in 90-day windows. Pull request #4504
- Flow direction filters on StarRocks deployments now work, fixing broken Internal/Outbound/Inbound/External chips. Pull request #4503
- StarRocks schema now migrates automatically at core startup instead of requiring manual DDL application. Pull request #4494
- V1.4.68 released with stateful alert rule events now carrying affected device, triggering event details, and a link back to the triggering event. Release
- V1.4.67 released with new CNPG clusters bootstrapping their database and core health events no longer reporting hourly healthy-to-healthy changes. Release
- NetFlow summary computation refactored to be testable without a database, enabling unit testing of previously database-only code. Pull request #4500
- NetFlow geo heatmap on StarRocks deployments now resolves country breakdowns through the GeoIP cache instead of failing on unsupported fields. Pull request #4497
Why it matters
StarRocks integration is now production-ready for NetFlow and telemetry, automatically handling schema setup and fixing critical query failures that blocked deployments. Core performance on the Network Flows page improved dramatically by collapsing serial queries, and map visualizations now show complete data instead of truncating to arbitrary limits.
Changelog entry
- Analytics: StarRocks warehouse now serves NetFlow and telemetry data as opt-in alternative to CNPG Pull request #4485
- Web: Network Flows page collapsed 14 serial loaders into 1 query Pull request #4499
- Core: Topology graph replaced with Dgraph, keeping AGE as shadow during rollout Pull request #4490
- Web: Dashboard NetFlow map now renders every arc instead of top 26 Pull request #4504
- SRQL: Flow direction filters now work on StarRocks deployments Pull request #4503
- Analytics: StarRocks schema auto-migrates at core startup Pull request #4494
- Analytics: NetFlow country breakdowns on StarRocks use GeoIP cache Pull request #4497
- Web: Sankey top edges now ranked over full window instead of 200-row limit Pull request #4502
- Web: Dashboard map draws every geolocated place reached by NetFlow Pull request #4501
- V1.4.68: Stateful alert rule events carry affected device and link to triggering event Release
- V1.4.67: New CNPG clusters bootstrap database; core health events stop reporting hourly healthy-to-healthy changes Release
V1.4.67–68 shipped. StarRocks now serves NetFlow and telemetry. Network Flows page 14× faster. Dashboard maps show all arcs. Flow direction queries work on StarRocks. Schema auto-migrates. Alerts carry device context.
V1.4.67–68 brings StarRocks to production for NetFlow and telemetry warehousing with automatic schema migration. The Network Flows page went from 14 serial queries to 1, fixing perceived slowness on fast warehouses. Dashboard maps now render complete data instead of truncated samples. Flow direction filters and country breakdowns work on StarRocks deployments. Alert events now carry device context and links to triggering events.