Documentation

Featurebase 4.x

Release notes - Molecula - Version 4.8.5

Warning

  • If you have initialized a data directory using 4.8.3 or prior, you will encounter the following issue when upgrading to 4.8.4: running server: opening server: opening Holder: open index: name=index_name, err=opening fields: decoding create field message: unmarshaling: unmarshaling CreateFieldMessage: proto: wrong wireType = 0 for field Gob

  • The solution is download the new version (4.8.5). Do not try to use 4.8.4 - it is not a production release and has been removed.

Bugs

  • FB-1487 Fix datagen no index error to specify the correct flag
  • FB-1489 Reduce memory usage of UnionRows
  • FB-766 Improved handling of queries containing unknown fields

Reversion

  • This fix was removed due to backwards incompatability - the issue will be reopened
  • FB-1359 DecimalAgg computations can have rounding error

Release Notes - Molecula - Version 4.8.4

Bugs

  • FB-1478 Handle nils in timestamp
  • FB-1471 Improve error messaging for checking value within range
  • FB-1381 RBF max file size can actually be hit
  • FB-1359 DecimalAgg computations can have rounding error
  • FB-1085 Address _keys field being created in roaring-migrate
  • FB-430 IDK - Panic when string data has a comma

Task

  • FB-1334 Add the ability to update fields to Null via the Kafka Static Consumer

Release notes - Molecula - Version 4.8.3

Bugs

  • FB-1468 Molecula-consumer-csv ingest yields inconsistent ingested data
  • FB-1462 Custom Epoch for Timestamp are being ignored
  • FB-1393 Improve handling of bad import values on ingest
  • FB-237 Generate an error for queries done on field that does not exist

Task

  • FB-1379 Create a featurebase subcommand to obtain an auth token
  • FB-1302 Create an endpoint to ingest all data for a shard transactionally

Release notes - Molecula - Version 4.8.2

Bugs

  • FB-1440 Fixed an Avro ingest issue that was caused by an overflowed integer value
  • FB-1456 Fixed TTL deleted views not returning correct results
  • FB-1436 Fixed an error where an out of range decimal fields return 0 instead of nil
  • FB-1435 Fixed an error where int fields with a positive lower bound affect the returned value
  • FB-1393 Improved handling of bad/unexpected import values on ingest
  • FB-1323 Fixed TTL Removal produces invalid time quantum state

Changes

  • FB-1378 IDK now uses whitelisted IPs for auth with FeatureBase when Auth is enabled
  • FB-1377 Grafana plugin now uses refreshed tokens provided by FeatureBase when Auth is enabled

Additions

  • FB-1459 Support explicit list of items in Rows call (‘in’ argument)
  • FB-1400 Improve Acceptance Criteria and Testing Plan for View TTL Feature

Release notes - Molecula - Version 4.8.1

Bugs

  • FB-1401 Remove check against etcd when executing queries - no impact to a cluster in a normal state, if cluster has one or more down nodes the query timeout will be the max
  • FB-1402 Remove check against etcd when executing backups - backups now work even in cases where etcd believes the cluster health is degraded
  • FB-1406 There is now a HeaderTimeout flag for the FeatureBase Backup command to set the HTTP client timeout, accepts an argument such as ’10s’.
  • FB-1416 molecula-consumer-csv now allows ingesting null Timestamp values
  • FB-1433 Kafka consumer now commits the offset correctly for the last message in a partition

Changes

  • FB-1384 Kafka consumer no longer auto-committing messages (available in v4.8.0)
  • FB-1387 In the Grafana plug-in, filtering can now be disabled for high-cardinality fields
  • FB-1390 New feature flag that enables a different, more balanced, shard distribution scheme. This is potentially destructive and cannot be undone. See documentation for full details.
  • FB-1238 Added support for Timestamp field queries from python client

Additions

  • FB-1399 TTL can now be implemented on any existing field that has a time quantum

Release notes - Molecula - Version 4.8.0

Bugs

  • AuthN/Z improvements
  • FeatureBase now iterates over all groups in Azure Active Directory when validating group membership for AuthN/Z
  • Set Authorization header appropriately when in a multi-node environment
  • Fixed bug that could cause deadlocking on ingest API delete call
  • Fixed bug when parsing To and From Filters on Rows with Time Quantum fields
  • Fixed bug where bitmap length 0 caused FeatureBase to crash
  • Fixed bug where Distinct() call did not work for Timestamp field
  • Fixed bug where GroupBy could return no results when time quantum fields and non-time-quantum fields were in the same GroupBy
  • Fixed bug where the kafka consumer was auto-committing offsets instead of only committing after they were confirmed ingested
  • Increased timeout window for nodes to be declared unhealthy under heavy load
  • Improved error messaging for SQL query on FeatureBase index that does not exist
  • Resolved issue with shards becoming unavailable under significant load
  • Standardized handling of time ranges across different calls
  • Timestamp fields now apply valid defaults if the provided min/max values are 0
  • Timestamp FieldOptions now includes Min and Max when marshalled to JSON
  • TranslateData now redirects to a healthy node only if local node is in an unknown state
  • Vulnerability fixes
  • Mitigated one instance of Integer Overflow or Wraparound
  • Mitigated one instance of Potential Integer overflow made by strconv.Atoi result conversion to int16/32
  • Mitigated two instances of Use of Hard-coded Password

Changes

  • Built all IDK artifacts for linux/ARM64, linked dynamically, and including molecula-consumer-kafka-sasl
  • FeatureBase new allows disabling fsync() for the RBF WAL only
  • Standardized commit timeout in IDK to ten seconds

Additions

  • New Delete() functionality
  • Added ability to free up disk space by deleting records from FeatureBase
  • Keys are now reused after their records are deleted

Documentation

  • Documentation on new Delete() functionality is available Here

Release notes - Molecula - Version 4.7.6

Bugs

  • Fix crash when Row call with “==” in Delete
  • Fix an issue where ingest could fail with an invalid page type error.
  • Grafana: fix an issue where grpc message size could be too large.

Additions

  • Grafana: allow disabling of a variable so not to send large queries

Release notes - Molecula - Version 4.7.5

Bug

  • Fix auth token retrieval so that we can use IDK with external-generate
  • Fix holder FlushCaches does not notice when indexes are being closed or deleted

Changes

  • Remove log on Kafka consumer unknown event
  • Upgrade all builds to Go 1.17.8 due to CVE in 1.17.7

Release notes - Molecula - Version 4.7.4

Bug

  • Fix a bug in which molecula-consumer-kafka is not respecting offsets
  • Fix an issue which prevented ingesting with –external-generate when Auth is enabled [a bug relating to obtaining Auth tokens was detected during this process and will be fixed in v4.7.5]
  • Grafana: Handle ExternalLookup() when using Time Filter [this item was mistakenly included in the v4.7.3 release notes, but it is actually available in v4.7.4]

Changes

  • None in this release

Release notes - Molecula - Version 4.7.3

Bug

  • Fix bug where creating a field that already existed did not return the correct error code, causing ingesters to believe that field creation had failed.
  • Fix a number of bugs with decimal fields and min/max values being out of range for a given scale due to JSON floating point rounding.
  • Fix bug where roaring-migrate would set min and max of decimal fields to zero.
  • Fix a number of bugs around Kafka and authentication.
  • Grafana: Handle ExternalLookup() when using Time Filter
  • Fix bug where auto-quoting in Web UI produced invalid SQL
  • Fix bug where field creation in a cluster could potentially deadlock
  • Fix bug where SQL parser panics when unknown field in GROUP BY
  • Fix issues with SQL Timestamp Parsing
  • Fix Corrupt RBF data when using pql Delete
  • Fix issue where ingest Using External Generate w/ TLS produces Error
  • Fix bug where RBF Close doesn’t wait for existing read Tx to close
  • Fix CSV consumer logs an error in a different format than expected
  • Fix panic over invalid date/time in query
  • Improve error message when Rows() “like” argument used on non-keyed set field
  • IDK allow space between configured primary key fields
  • Fix Kafka-consumer doesn’t handle kafka host name correctly

Changes

  • Tweaks to release structure based on feedback. Include version numbers for all artifacts.
  • Remove ui/usage and cardinality (schema/details) support. This is removed due to causing significant performance problems.
  • Change release notes to text format. PDF version of past release notes is included, new release notes will be prepended to this file going forward.

Additions

  • Make config options consistent for Kafka SASL
  • Implement TTL on time views
  • Enhance Deletion to include key-id mappings and “reclaim” deleted records
  • Upgrade go to 1.17.7
  • Enhance SQL functionality:
    • DISTINCT queries now support the WHERE clause
    • GROUP BY queries now support “AND” & “OR” in the WHERE clause
  • Improve groupby/aggregate performance by pooling containerFilter objects

Testing

  • Extra tests for etcd retry code
  • Improve RBF delete test coverage

Documentation

  • Add docs around making sure atime is off and fstrim is on

Release notes - Molecula - Version 4.7.1

Grafana Plugin update

  • Add new functionality and ability to use ID values for variables in queries
  • Create ability to add Integer values from variables in queries
  • Support multi-valued variables capabilities
  • Health check for server before submitting a query

FeatureBase update

  • Enable refresh token capability to the FeatureBase Server
  • Added compatibility with Confluent Cloud for Kafka-Avro-SASL compatible IDK
  • Support on AMD64 and ARM64

Bug Fixes

  • Remove segfault errors for on empty row in timestamp
  • Fix cache invalidation issues caused by Backup and Restore

Release notes - Molecula - Version 4.7.0

Bugs (Fixed)

  • Decimal value doesn’t always match source (aka pre-ingest) value
  • GRPCurl GetIndexes command throws error
  • User with ReadOnly access is able to drop table
  • Web UI Table Page 404 error loading
  • Really slow imports on int or mutex fields if records span multiple shards and unsorted
  • Panic when sending an http “POST /transaction” to a non-primary node
  • Concurrent map access during Restore
  • “featurebase rbf pages” errors
  • Panic during setRow and subsequent panic on restart
  • Fatal error: fault under concurrent query load
  • Panic in roaring bitmapToArray when running difference queries
  • Grafana - Query Building - Extract(Limit())
  • Grafana - Query Building - Count(Distinct())
  • Fix typo “translaste” in roaring-migrate
  • Don’t panic on non-existent translate store
  • Backup not getting all index keys
  • Backup/Restore does NOT either backup or restore int and decimal data types
  • Issue with RBF WAL size check and WAL related configuration

Story

  • Import PQL parser and rewrite dynamic queries to use it
  • Add result limit option to Grafana plugin
  • FeatureBase tooling Authn/Authz enablement
  • Protect GRPC Endpoints
  • Filter response of HTTP endpoints that return holder-level information
  • Percentile queries on all int-like fields
  • Grafana Plugin - Add Ability to Authorize User
  • Update featurebase.conf to enable configuring an arbitrary OAuth2.0 provider as the Identity Provider for FB AuthN/AuthZ
  • Authentication in the FeatureBase UI
  • Ability to map user groups to cluster admin / index-level permissions
  • Check if AuthN/AuthZ is enabled in handlers

Task

  • FeatureBase restore command emits extra-verbose messages
  • Secure Intra-cluster communication
  • Expose etcd.dir config option to allow user to separate FeatureBase and etcd I/O
  • Stop relying on cluster status for query validation
  • Ensure FeatureBase logs rich enough data about requests when AuthN/Z is enabled
  • Check user authorization on all HTTP endpoints that access indexes
  • Queries taking way longer under ingest load, but for the wrong reasons
  • featurebase.conf, featurebase.service and ReadMe.md added to git/releases

Release notes - Molecula - Version 4.6.2

Bug

  • Fix build of molecula-consumer-sql

Release notes - Molecula - Version 4.6.1

Bug

  • Featurebase import error using RBF for a field that was deleted and recreated roaring-migrate fails when source directory has trailing slash RBF data corruption.
  • Kafka-Avro error with Avro Schema
  • Errors during streaming result get mishandled and lost
  • Count on Distinct() on timestamp field broken
  • Add new RecordBatch implementation which uses ingest API
  • Perform sanity check only if we have the storage it runs on
  • Fix bug where timestamp val is wrong in Min/Max
  • Fix presentation of timestamps in Groupby, Distinct calls from psql client Fix featurebase holder subcommand
  • Use go1.16.10 for build & test
  • Fix –string-array-separator flag in IDK

Task

  • Allow TopK queries on mutex fields

Release notes - Molecula - Version 4.6

Bug

  • Runtime error: index out of range
  • Cluster State Sync - remaining time estimate is way off Deleting and re-creating a field can cause deadlock
  • Refactoring
  • go vet / staticcheck issues

Story

  • Grafana plugin needs timestamp mapping
  • rework import to not need to talk to the local node over the network
  • Remove stale Tx/Qcx API parts
  • Add logging to Roaring migration tool to show progress
  • Execute SELECT … INNER JOIN statements and return a result set Ability to Backup and Restore a single FeatureTable from a FeatureBase Continuous Delivery of Changelog
  • CSV ingest - out of range error message needs more info

Task

  • /schema/details can bog down FeatureBase on some data sets Package relevant FeatureBase binaries as Linux packages (deb, rpm) Improve testing for background key replication
  • Kafka Consumer - Need ability to Auth using SASL

Pilosa v3.6.0

Changes since v3.5.0

  • Rework Import to reduce copying and network traffic
  • Add deb/rpm packaging of FeatureBase binary
  • Slow webui on large query
  • Create FeatureBase Docker Images in CI
  • Add quotes around strings in webui representation
  • uprev gopsutil
  • Remove shards list from error message entirely
  • Remove bolt backend, bluegreentx, and a ton of unused API surface
  • Aggregate COUNT() with INNER JOIN
  • Fsync
  • Test paused node picks up once cluster state is back to normal
  • remove Pilosa name and website from curl handler for webUI endpoints
  • Allow StmtRows.Scan() for more types

    Added

  • FB-900: aws single node deployment for linux amd64 binary
  • add fgprof endpoint by default
  • Add cmd option to disable cardinality calculation in schema/details endpoint
  • Migrated Cloud Build to GitLab
  • Added in various CI files
  • Partial backup/restore
  • Featurebase import: check for data in requests
  • wrap shards list in ShardSlice for prettier output in error messages
  • Change time estimation to use avg time per message
  • Fix deadlock on field recreation after node restart
  • adds more detail to CSV ingest error message
  • don’t close storage after failing to open cache

    Performance

  • don’t fsync on RBF Open if WAL is empty
  • rework executor’s per-shard union to use UnionInPlace

    Changed

  • Add darwin build to roaring-migrate-tool

IDK v0.24.0

  • remove buggy test
  • kafka-sasl ingest retrieve all records from beginning when SkipOld is false
  • add some conditions to properly install on an M1 mac
  • add context to query error message
  • Move 4MB of false positives into .data.go files to enable filtering

    Changed

  • Switch postgres lookup table ID to bigint

    Added

  • Added readme for Kafka SASL
  • Install ca-certificates in docker
  • Added Kafka-SASL final
  • CKafka ssl
  • Update TLS stuff to support reading cert data from environment vars
  • Read kafka-static schema from an S3 bucket

Grafana Plugin v1.0.4

Changes since v1.0.3

Added

  • Grafana plugin needs timestamp mapping

Molecula v4.5.1 2021-09-30

IDK v0.23.0

Changes since v0.22.0

Added

  • Build linux-arm releases

Molecula v4.5.0 2021-09-28

FeatureBase v3.5.0

Changes since v3.4.0

Added

  • Cluster support for ingest API
  • Looker Integration
  • Add SQL type checker
  • Handle SQL comments during scan
  • Implement SQL GROUP BY (in new engine)
  • Added DecimalAgg field in GroupCount to output decimal sum
  • Add support for ARM64
  • Implement non-aggregate SELECT query (in new engine)
  • Add mutex field sanity-check endpoint
  • Handle SQL WHERE clause (in new engine)
  • Implement basic SQL COUNT(*) query (in new engine)

Fixed

  • Better error message for ‘cannot allocate memory’ error
  • Fix several issues uncovered in mutex fields that could allow multiple values to set erroneously
  • Replicate keys to replica nodes which are up even if some are down

Changed

  • Rename docker builds to featurebase (with docker fix)

IDK v0.22.0

Changes since v0.21.0

Added

  • Add schema subject and version to log message for Kafka ingester

Molecula v4.4.0 2021-08-17

Pilosa v3.4.0

Changes since v3.3.0

Fixed

  • Differentiate error messages between opening file and mapping it
  • Fix data race in batch import client by adding locking around rowCache
  • Wrap field and table names in backticks for sql queries

Changed

  • Disable /ui/usage endpoint completely when usage-duty-cycle is set to 0
  • Rename pilosa binary to featurebase
  • Add logging for and allow disabling usageCache
  • Update metric names to use “featurebase” prefix

Added

  • (UI) Add aggregate sort option to query builder
  • Add max memory limit to Extract() to prevent OOM
  • Migration tool for RBF
  • FeatureBase Renaming
  • Add feature flag –future.rename to support rename to FeatureBase
  • (UI) Add ability to sort when building GroupBy queries

IDK v0.21.0

Changes since v0.20.0

Added

  • FeatureBase Renaming: IDK commandline flags
  • Add feature flag –future.rename to support rename to FeatureBase

Grafana Plugin v1.0.3

Changes since v1.0.0

Added

  • Fix issue where list of indexes wouldn’t populate until swithcing from PQL to SQL and back
  • Automate builds & release for grafana plugin

Pilosa v3.3.0

Changes since v3.2.1

Added

  • CI test for backup restore
  • Add “pilosa chksum” sub-command for easier data validation
  • Parallelize restore command
  • Add ‘last updated’ info to tables UI
  • Add UI support for single codepoint ‘_’ matches in LIKE queries
  • Allow all field types for query builder in UI
  • Allow like for query builder queries
  • Add additional logging for LookupDB connection
  • Multi-node restore
  • Add an option to open a transaction in an external lookup query to allow deleting from Postgres
  • Allow filter parameter in query builder (UI) GroupBy
  • Add darwin-arm64 support
  • Add support to restore from backup
  • Add support for backups from clusters

Fixed

  • Fix bug where deleting an index didn’t delete all caches
  • Fix bug where missing ‘field’ argument on Percentile query would cause a crash
  • Fix lack of support for timestamps when using Postgres wire protocol
  • UI: Fix issue with non-applicable saved queries being shown in GroupBy’s filter list in QueryBuilder
  • Fix Percentile bug caused by integer overflow when dealing with very large values
  • Fix issue where LIKE filter might not match recently created keys
  • Fix 2 cases (Not and time quantum queries) of mixing write operations (Store) with reads in a single request causing deadlocks Performance
  • Improve key translation performance by only using the newer find and create methods. Add wrappers for old methods for compatibility
  • Improve importValue performance, fix potential bug with unstable sort

Changed

  • Cache results of ui/usage endpoint with configurable update interval to avoid slow concurrent usage calculations
  • Remove redundant shard list from error message
  • Use etcd to maintain a consistent list of available shards rather than local file

Removed

  • Remove ProdPad integration from UI
  • Remove support for attributes in Pilosa due to performance/scaling issues and lack of use

IDK v0.20.0

Changes since v0.19.0

Fixed

  • CSV Ingester: If row has a different schema than header, raise an error, dont crash
  • Upgrade Go to 1.16 for builds because of darwin/arm
  • Fix bug where DateInt field’s time unit configuration could not be read
  • Fix libc issue in ODBC binaries by building on Ubuntu instead of Alpine

Changed

  • Use storage API for BigQuery ingester for better performance

Added

  • Add builds for darwin-arm64

Molecula v4.2.1 2021-05-10

Packages included:

  • Molecula Pilosa v3.2.1
  • Molecula Consumer v0.19.0
  • Python-Molecula 1.8.0

Summary:

This patch release includes a number of bugfixes and performance improvements, and a large amount of new UI functionality. As noted in the changelog, it also includes a prototype for transactional backups which is not yet completely functional and should not be used.

Changes in this release:

Pilosa v3.2.1

Changes since v3.2.0

Added

  • Update to latest UI including Lookup functionality
  • Backup CLI Prototype

Performance

  • Improve performance on mutex fields with sparse writes
  • Make key creation faster for large batches

Fixed

  • Error instead of panic on “/inspect” endpoint
  • Fix RBF root record cache build
  • Avoid nil pointer exception when failing to get listener

IDK v0.19.0

Changes since v0.18.0

Fixed

  • Delete spamming print statement for timestamp
  • Fix bug where number of record fields was computed incorrectly

Molecula v4.2.0 2021-04-30

Packages included:

  • Molecula Pilosa v3.2.0
  • Molecula Consumer v0.18.0
  • Python-Molecula 1.8.0

Summary:

This release includes several major features including timestamp fields, the Delete query, and more!

Changes in this release:

Pilosa v3.2.0

Changes since v3.1.2

Changed

  • Deprecate ‘runtime’ key on query-history response and add ‘runtimeNanoseconds’
  • Add log prefix levels

Fixed

  • Improve setting/checking of file/mmap limits, don’t attempt to set hard limits which requires root
  • Correct incorrect bit counts and related errors from pilosa inspect command (roaring backend)
  • Correct/eliminate spurious diagnostic messages about memory ownership or corruption
  • Eliminate spurious ops log entries in roaring backend for deleting bits which were already absent (does not affect RBF)
  • Fix potential data loss bug in edge case in an importValue path
  • etcd: overhaul interactions to improve performance and reliability using heartbeats and caching

Added

  • Add PQL Delete function for deleting all records selected by a query
  • Add Timestamp field type

IDK v0.18.0

Changes since v0.17.1

Added

  • Add prototype of HTTP/JSON ingest API
  • Support new timestamp field in Kafka ingester
  • Add Lookup Text Field, with postgres backend
  • Add BigQuery ingester
  • Implement offset-based IDA location within IDK
  • Allow RecordTimeField to use epoch dates as well as datetime strings
  • Add “AllowMissingFields” flag in Kafka Static

Molecula v4.1.1 2021-03-15

Packages included:

  • Molecula Pilosa v3.1.1
  • Molecula Consumer v0.17.1
  • Python-Molecula 1.8.0

Summary:

This patch release includes several usability updates, bug fixes, and performance improvements in both Pilosa and the Consumers. Changes in this release:

Pilosa v3.1.1

Changes since v3.1.0

Fixed

  • Fix bug where integer fields wouldn’t get loaded properly on restart
  • Fix IDal location API which was incorrectly denying requests
  • More graceful error handling for corrupt container data on import
  • Change protobuf package version to suppress deprecation message about “github.com/golang/protobuf/protoc-gen-go/generator” package
  • Create the data directory if it doesn’t already exist to avoid error log
  • Represent zero indexes in /schemaresponseas[] instead of null
  • Check for nil translates to re while reading translate entries to avoid crash
  • Ensure cluster/node metadata sharing is consistent in etcd

    Added

  • Add internal testing “oracle” to check random data/query results against alternate implementation

IDK v0.17.1

Changes since v0.17.0

Performance

  • Improve performance of batch imports by de-nesting internal maps

Fixed

  • Change “coordinator” error messages to “primary node”

Molecula v4.1.0 2021-03-08

Packages included:

  • Molecula Pilosa v3.1.0
  • Molecula Consumer v0.17.0
  • Python-Molecula 1.8.0

Summary:

This is the second release in the Molecula 4.x series! It includes another breaking change to the data directory structure in Pilosa, and so 4.0.0 should not be used. It also includes a revamped molecula-consumer-kafka-static which supports nested JSON.

Changes in this release:

Pilosa v3.1.0

Changes since v3.0.0

Changed

  • Remove stutter and suffix from backend files

IDK v0.17.0

Changes since v0.16.0

Changed

  • Implement a more flexible kafka static schema. This changes the kafka-static ingester to use a JSON-based schema specification, and supports nested JSON.

Added

  • Add key translate batch size flag

Molecula v4.0.0 2021-03-05

Packages included:

  • Molecula Pilosa v3.0.0
  • Molecula Consumer v0.16.0
  • Python-Molecula 1.8.0

Summary:

This is the first release in the Molecula 4.x series! It includes a number of major improvements and also breaks backward compatibility in terms of the data format and configuration file for Pilosa.

The highlights are:

  • The new RBF storage backend which provides:
    • ACID semantics at a shard level.
    • Massively reduced heap usage for more predictable memory consumption and garbage collection behavior.
    • Ability to scale a single node far beyond the size of its memory.
  • Etcd for clustering and node membership. Pilosa 3.0 embeds etcd which uses a Raft-based consensus algorithm to power a strongly consistent key/value store; Pilosa uses this to maintain cluster state, node state, and data schema. This reduces internal complexity in Pilosa’s codebase while providing a more reliable store for important metadata.

Changes in this release:

Pilosa v3.0.0

Changes since v2.9.0

Added

  • Percentiles query Changed
  • Reorganize Pilosa data directory to avoid naming conflicts and support multiple storage backends
  • Replace memberlist/gossip with etcd/Raft
  • Add RBF storage backend as the newd efault; adds ACID semantics and low-level transactional support

Fixed

  • Fix bug where cancelling a query might orphan workers
  • Fix a racec ondition when deferring foreign-index

initialization

  • Defer cluster messages until startup
  • Fix enabled WebUI url log
  • Fix GroupBy test which uses RowKey insteadof RowID
  • Global Port Mapper avoids many races in portal location for cluster setup
  • Avoid precondition failed (412) on ingest

Removed

  • Removed ocs directory (nolongerused)

IDK v0.16.0

Changes since v0.15.0

Changed

  • Updated to support Pilosa3.x series (for Molecula 4.x)
Something missing or incorrect?
Help improve this article or join us on Discord!

© 2022 Molecula Corp. (DBA FeatureBase). All rights reserved.