Skip to content

Latest commit

 

History

History
210 lines (162 loc) · 9.34 KB

CHANGELOG.md

File metadata and controls

210 lines (162 loc) · 9.34 KB
Raw

Changelog

1.7.3 - 2020-09-21

Added

  • Add options to choose host selection policy.
  • Add option to use server-side timestamps for queries.
  • Add support for password authentication.

Fixed

  • Don't enable tracing if it's not requested.

1.7.2

  • schema: add "IF NOT EXISTS" check when creating a new type

1.6.9

  • Avoid DDL operations for a table with MV (#198)

1.6.1

  • Add support for generating multiple tables in schema with the --max-tables command line option.

1.6.0 - 2019-09-06

  • Bumped driver version to v1.3.0-rc.1
  • Lazy partition key generation reintroduced to avoid out of memory issues. This brings in a new CLI arg --token-range-slices that defines how many slices the partition keyspace should be divided into when applying the different distribution functions. The default value of this is set to an ad-hoc value of 10000 which should supply ample possibilities for varying selection of values according to the chosen probability distribution.

1.5.0 - 2019-08-26

  • Fix overlapping operations on the same partition key (#198).
  • Partition keys can now be drawn from various distributions such as "zipf", "uniform" and "normal". The CLI argument --partition-key-distribution is used to select which distribution to use. The default is normal.
  • The CLI argument partition-key-buffer-size is removed since it carries no meaning any more.
  • Added the possibility to run without any validations against the Oracle. Simply do not supply a host for the Oracle and Gemini will assume you want to only run against Test.
  • Replication strategy is now configurable via the CLI argument --replication-strategy.

1.4.4

  • Mutations on Test are only applied if they first succeeded on the Oracle.

1.4.3

  • Bugfix that makes sure that when a job terminates early, the result status is properly sent to the collector.
  • Gemini ensures that material views can be created in the default case by simply creating enough keys and columns.

1.4.2

  • Reused primary keys does no longer block the caller if none are available.
  • Primary key generation no longer blocks if the targeted source is full.
  • Upgraded driver to 1.2.0

v1.4.1

  • Bug in shutdown handling that caused deadlock is fixed.
  • Index queries reapplied with low frequency for certain types.
  • Fix for invalid materialized view ddl statement.

1.4.0

  • A source concept is used to coordinate the creation, consumption and reuse of partition keys.
  • Two new CLI args are introduced to control the buffer sizes of the new and reusable partition keys partition-key-buffer-size and partition-key-buffer-reuse-size.
  • The CLI arg concurrency now means the total number of actors per job type. You may need to scale down your settings for this argument since for example a mixed mode execution will run with twice as many goroutines. Experimentation is encouraged since a high number will also yield much greater throughput.

1.3.4

  • Shutdown is no longer waiting for the warmup phase to complete.
  • Partition keys can now be any supported type.
  • The size of the partition key buffers can be configured on the commandline through partition-key-buffer-size and partition-key-buffer-reuse-size.

1.3.0

  • Partitioning bewteen workers are now handled by a generator that generates random partition keys and dispatches them to the relevant worker. This construct also provides a channel for "old" values to be used downstream by validators. The reason for this is that if the validators read newly created partition keys then the chances of finding data in the database is near to zero unless we load it up with a truly huge data set.
  • Log levels introduced and configured via CLI arg level. Common values such as info, debug and error are supported.
  • Mutations are now retried at a considerably greater number of times than reads. The number of retries and the time between them are configure via --max-mutation-retries and --max-mutation-retries-backoff.

1.2.0 - 2019-06-20

  • DDL statements are now emitted with low frequency if the --cql-features is set to at least "all" level.
  • Data sizes are configurable though a CLI argument --dataset-size and the currently supported values are "small" and "large".
  • CLI toggle --cql-features added to let the user select which type of CQL features to use. The current levels are basic, normal and all. The basic level have only regular columns and no indexes nor materialized views. normal adds these two constructs and all currently the same as normal but will be used to differentiate more advanced features in the future.

1.1.0 - 2019-06-11

  • Exponential backoff retry policy added with 5 retries between 1 and 10 seconds.
  • Support for changing consistency level via a CLI argument consistency.
  • Support for compaction strategies added via a CLI argument compaction-strategy as a set of string values "stcs", "twcs" or "lcs" which will make Gemini choose the default values for the properties of the respective compaction strategies. Alternatively the JSON-like definition of the compaction-strategy can be supplied in a form like: {"class"="SizeTieredCompactionStrategy", "enabled"=true, ....}. Note that the form needs to be given as actual valid JSON.
  • Prometheus metrics added that exposes internal runtime properties as well as counts fo CQL operations 'batch', 'delete', 'insert', 'select' and 'update'.
  • Warmup duration added during which only inserts are performed.
  • Generating valid single index queries when a complex primary key is used.
  • Gracefully stopping on sigint and sigterm.
  • JSON marshalling of Schema fixed. The schema input file has changed to ensure marshalling of lists and sets. These types now have a kind property with possible values (list,set).
  • Correctly pretty printing map and UDT types.
  • Skipping HTML escaping of resulting JSON.
  • Ensure proper termination when errors happen.
  • Fix mutation timestamps to match on system under test and test oracle.
  • Gemini now tries to perform mutation on both systems regardless of if one of them fail.
  • Gemini now timestamps errors for easier correlation.
  • A new Store abstraction is introduced in preparation to enable implementations such as an in-memory store.
  • Gemini now uses github.com/scylladb/gocqlx/qb builder.
  • Gemini ensures that primary key buckets do not overflow int32.
  • Gemini now accepts a list of node host names or IPs for the test and Oracle clusters.
  • Default maximum primary keys increased to MAX_INT32/concurrency.
  • Range tombstones are being generated.

1.0.0 - 2019-05-06

  • Gemini version is now available in the resulting output.
  • Materialized Views support.
  • Improved error handling in validation code.
  • Avoiding small double booking of write ops in case of mutation errors.
  • Printing executable CQL statements when logging errors or in verbose mode.
  • JSON schema definition file has simpler index definition.

0.9.2 - 2019-04-18

  • Support for User Defined Types (UDT) added for simple columns.
  • Support for collections such as sets, lists and maps added for simple columns.
  • Support for writing the result to a file in JSON format.

0.9.1 - 2019-04-11

Added

  • Tuple support added for simple columns.
  • Added version info printing using '--version' program argument.
  • CQL INSERT JSON statement support.

Fixed

  • Panic when --non-interactive command line option is passed (#69)

0.9.0 - 2019-04-03

Added

  • Support for queries using secondary indexes added.
  • Switched to using the upstream Go driver https://github.com/scylladb/gocql instead of the regular driver. The goal is performance gains by using that shard awareness feature as well as providing proper more real testing of the driver.
  • Support for additional types for the clustering keys. TimeUUID is used runtime to accomplish sorted UUIDs. Blob, text and varchar types are based on ksuid to be sortable.
  • Launcher script (scripts/gemini-launcher) that starts an Apache Cassandra node as the test oracle and a Scylla node as the system under test using Docker.
  • Schema generation support. Gemini generates a random schema unless user specifies one with the --schema command line option.

Changed

  • Improve progress indicator (#14).