0.5.7

• support for cassandra retries & configurable host selection policy
• better cassandra metrics
• end to end testing in CircleCI.
• docker entrypoint script now actually verifies it can hold a connection open for a few seconds.

0.5.6

• instrument cassandra errors & timeouts. update dashboard
• update gocql cassandra libray. no impact expected.
• better docs, especially around operations. include overview diagram
• make heap profile trigger more simple to use
• simplify docker stack: reuse existing images, remove elasticsearch
• dashboard: some better charts, fix some units

0.5.5

test. same as 0.5.4

0.5.4

index:

• various improvements, doc updates
• support pruning & deleting from index (causes a slight increased load on the cassandra backend)
• bugfix: abort if metricIndex failed to initialize #325

other:

• improvements to clustering, operations, cassandra docs and install guides. in particular deb install guide fixes
• filter metric search results by comparing from TS a07007a to lastUpdate field from metricdefinition

0.5.3

index:

• add cassandra backed metric index
• better metric names (+ update dashboard)
• add metric for items in ES index retry buffer (+ update dashboard)
• dashboard: show index&cassandra durations better
• fix memoryIdx bug (39bfcc8)
• cleanups (no impact)

doc:

• split up installation guide into 3 different guides, and improve them

config:

• fix index config validation (ca0ff63)
• configurable cassandra keyspace
• better configs for docker stack vs in package
• default consumer-fetch-min = 1
• make statsd instrumentation optional, e.g. remove depedency on statsd.
• set consumer-fetch-default to sarama default of 32768
• fix bug that was resetting consumer-fetch-default to 100

core:

• fix null checking in series merging (e22236e)
• move kafkaMdm input and kafka cluster plugin to manage their own offsets, instead of using sarama-cluster for consumergroups which was pointless.
• display request information when it causes a panic.

0.5.2

new metadata index

• refactored metadata& caching into new index package
• provide memory and elasticsearch backend

core:

• allow adding points to the latest, but already saved chunk
• various persister changes. #263
• fix carbon input which was broken since new metrics2.0 version.
• configurable http limits. fix #268
• add ES index fail metrics
• truly validate metrics in ingest

builds & packaging:

• refactor circleCI building
• (graphite-raintank is now graphite-metrictank)
• fix packages for centos 6 & 7, debian/jessie
• add a docker stack
• auto push docker image to dockerhub when circleci builds
• switch to AGPL
• include include storage-schemas.conf in packages

docs:

• various docs update, put docs table in readme
• simplify readme
• add metrictank dashboard, and add on grafana.net

0.5.1

• rename metric_tank/metrics_tank to metrictank. this changes all the statsd metrics(!)
• add kafka clustering transport next to nsq
• upgrade to new, proper metrics2.0 schema, which changes how metricId's are generated, but ...
• ....support series merging to seamlessly combine data for the same metric living under different metricdefs ff7a9bc
• instrument input decoding errors
• adopt govendor for go vendoring #197

config:

• proper sample config that actually makes sense
• make configuration of input plugins more elegant
• kafka offset: default to newest (#236)
• switch fallback schema from DateTieredCompactionStrategy to TimeWindowCompactionStrategy #249
• configurable CQL proto version
• kafka-mdm: expose performance tuneables

doc & packaging:

• various bugfixes in packaging scripts and packages.
• build both ubuntu14.04 and 16.04 packages, and centos 6 & 7
• doc updates, especially around "data knobs", usage reporting, consolidation, http api, cassandra, index plugin writing, clustering and readme.

0.5.0

This is the first "release", and is the culmination of a lot of work.
Some highlights of recent developments leading up to this version:

• rename metric-tank -> metrictank
• promote metrictank to top level of repo and move other stuff out, start versioning metric schema's in other repo
• add kafka-mdm, kafka-mdam and carbon input next to nsq.
• accept empty inputs instead of calling log.Fatal #226
• don't busy loop and max out cpu if metricpersist production has issues
• make sure metricpersist msgs can't go over nsq's max size. #224