1.0.2: rolling upgrade fix

Overview

This release fixes module work for rolling upgrade over existing cluster without state space (migrations 0.x or clean cluster).

Fixed

• State space bootstrap on cluster rolling upgrade (gh-77)

1.0.1: license introduction

Overview

This release adds license to the module.

Added

• BSD License.

1.0.0: per-replicaset state

Overview

This release introduces new approach to storing migration state: per-replicaset instead of clusterwide. New approach fixes the issue when new replicaset did not applied the migrations that are already applied on the remaining part of cluster.

To start using 1.0.0 instead of 0.x.x module on existing application, one must first call move_migrations_state through HTTP or Lua API to move current cluster migration state from Cartridge clusterwide configuration to per-replicaset utility spaces. See README.md for upgrade and downgrade scenarios.

Fixed

• Migrations do not apply to newly added replica sets in the cluster (gh-65). Applied migration names are moved from the cluster-wide configuration to the space on each node.

Added

• An API for moving existing migration names from the cluster configuration to a space.
• API for getting applied migrations list for the cluster.

0.7.0: check enabled roles

Overview

This release introduces new helper to check whether roles are enabled on the Cartridge instance

New features

• utils.check_roles_enabled helper function to check whether roles are enabled on the instance (#68).

Thanks

We want to thank @Satbek for a feature contribution.

0.6.0

Overview

This release introduces new API to configure timeout for storage migrations.

New features

• Configurable timeout for storage migrations (#66).

Infrastructure

• Fix running tests with Tarantool 2.11+ (PR #67).
• Fix running tests with tarantool/http 1.2.0+ (#63).

0.5.0

Overview

This release introduces new API to check module version in code.

New features

• Add versioning support (PR #62).

0.4.2

Fixed

• Fetch schema from a replicaset leader to apply on the clusterwide config even when migrations.up() is called on a replica (gh-56). The local schema on the replica may be not the most actual due to replication lag.
• Issue a warning into log when register_sharding_key() is called with {'bucket_id'} key (gh-49). It is likely a mistake: sharding key is a set of fields, which are used to calculate bucket_id, not the bucket_id itself.

0.4.1

Fixed:

• Unclear error output in some cases

0.4.0

Added:

• Lua API to trigger migrations from console

Fixed:

• Fix crash during init when instance http server disabled

0.3.1

Features

Added config-loader to load migrations from cartridge clusterwide config

Fixes

Fixed "fiber name is too long" error when running on instances having long instance name