added
24 September 2021 - v.2.18.5
about 3 years ago by Jason Turner
Download: hdxctl-v2.18.5
Release Location
wget -O hdxctl https://hdx-infrastructure.s3.amazonaws.com/hdxctl-v2.18.5 && chmod +x hdxctl
Features
Ingest
- Ingest: Add max_future_days limiting for tables. Enables the ability to filter date times that are presented too far in the future. Default is this is turned off.
- Ingest (Batch): Add additional information into the log to include the the input path.
- Ingest (Batch): Ensure Batch Peers always remove SQS entries from the queue.
- Ingest (Kafka): Add offset for Kafka peer readers.
Query
- Query: Enable ability to see active queries.
- Query: Add query server metrics to Prometheus
Portal/API
- Portal: Additional input validation and error handling.
- Portal: Enable Shard Key in Table settings.
- Portal: Add retry behaviour in Batch Jobs in the UI.
- Portal: Add delete stopped jobs from the UI.
- Portal: Add Kafka section to Cluster UI.
- API: Change behaviour of Transforms and views PATCH to PUT.
Observability
- Grafana: New Query Dashboard. Deployed in Grafana. Will be deployed with build later.
Bug Fixes
Ingest
- Ingest: Intake will no longer create SQS queues if they do not exist, instead logging an error.
- Ingest (Batch): Fix a goroutine leak when batch ingesting JSON that can exhaust memory.
Query
- Query: Fixed segfault in query engine where unindexed partitions were queried with Indexed ones.
- Query: Add partition option to query API (for debugging purposes).
- Query: HTTP client bug fix where at high load the HTTP client could fail when collecting S3.
- Query: Do not include internal queries in hdx.active_queries
Lifecycle
- Merge: Partitions are sorted by modification date (most recent first) prior to being provided to the Merge SDK. The Merge SDK will use the first partition as the prototype for what a merged partition's schema should look like.
- Merge: Bug fix where a non-indexed column is changed to a column caused the Merge-peer to crash.
- Merge: Resolve an issue where merge components were unable to locate the queue when running without merge pools