v1.2.9

API

• Set default API version to 1.30
• Fail volume create requests if they fail on any node
• Avoid false positive errors if container networkID is not set
• Drop support for old TLS versions
• Disallow passing HostConfig on container start
• --since and --before filtering for docker images
• Add support for the health and is-task filters
• Remove all usage of samalba/dockerclient
• Use http client created in Swarm
• Make sure we don't read from Request.Body after hijacking
• Add automatic constraint logic when ContainerCreate fails because of an OSType mismatch
• Support image load on hybrid cluster
• Only check healthy engines when doing image inspects
• Remove check against minimum version

Mesos

• Remove Mesos integration due to lack of maintainability and updates.

Discovery

• Fix No Elected Primary Cluster Manager issue

Stability

• Force resume refreshLoop when engine status becomes healthy
• Fix deadlock and address race condition in RefreshContainers
• Avoid concurrent map read-write panic
• Overhaul events code using watchQueues

Rescheduling

• Reschedule with new IP on disconnect errors

Miscellaneous

• Remove Windows binary

v1.2.8

API

• Support asterisks in image reference filter

Networking

• Fix panic when network is created without connecting to any containers

v1.2.8-rc1

Networking

• Fix panic when network is created without connecting to any containers

v1.2.7

API

• Add name filter for network ls
• Allow control over where volumes get created with a whitelist label
• Immediately flush response headers for ContainerWait requests
• Add support for network filters on containers and label filters on volumes
• Don't throw away most of the JSONMessage content when pulling images
• Add progress reporting during image pulls
• Add filter to list dangling images
• Add OSType as a label for node constraints
• Don't fail docker pull operations when running both Linux and Windows nodes
• Add support for some new image build options
• Node filters for networks and volumes

Events

• Move events handling to use watch queues from github.com/docker/swarmkit
• Event stream retry interval capped to 10 seconds

Networking

• Fix refresh loop after network ls stopped returning containers attached to networks (for newer API versions)

Discovery

• Deprecate Docker Hub discovery (token based); to be removed in an upcoming release

Misc

• Move vendoring to use github.com/LK4D4/vndr instead of Godeps
• Fix several test failures
• Parallelize engine operations

v1.2.6

API

• Add options to refresh engine container cache on filters
• Support name filter in volume list
• Support more filters in network list operations
• Support node whitelist filter
• Switch from engine-api to docker/api/type and docker/clients
• Pass size parameter through on container inspect

Scheduler

• For container network disconnect requests, try worker engine with the container first
• Precompile filter regular expression to reduce CPU usage
• Remove setTCPUserTimeout to avoid TCP connection leak
• Fix network endpoints for rescheduling

Cluster Management

• Remove dependency on IPv4 addresses
• Support event top, resize, commit and so on to avoid unnecessary refreshing
• Sequentialize event monitor to an engine to avoid data race
• When an active engine sends EOF on event stream, restart event monitor so we don't lose events
• When proxying a request, cancel request if user connection is broken
• Give up leadership when manager shuts down

MISC

• Update go-zookeeper to fix a lock spin problem
• Migrate documentation to https://github.com/docker/docker.github.io/tree/master/swarm
• Update Swarm CI to use go 1.7.1
• support GOARCH to be able to build for other architectures
• Send Swarm logs to stdout

1.2.6-rc2

API

• Support name filter in volume list
• Support more filters in network list operations
• Support node whitelist filter
• Switch from engine-api to docker/api/type and docker/clients
• Pass size parameter through on container inspect

Scheduler

• For container network disconnect requests, tryworker engine with the container first
• Precompile filter regular expression to reduce PU usage
• Remove setTCPUserTimeout to avoid TCP connection leak
• Fix network endpoints for rescheduling
• Give up leadership when manager shuts down

Cluster Management

• Support event top, resize, commit and so on to avoid unnecessary refreshing
• Sequentialize event monitor to an engine to avoid data race
• When an active engine sends EOF on event stream, restart event monitor so we don't lose events
• When proxying a request, cancel request if user connection is broken

MISC

• Update go-zookeeper to fix a lock spin problem
• Migrate documentation to https://github.com/docker/docker.github.io/tree/master/swarm
• Update Swarm CI to use go 1.7.1
• support GOARCH to be able to build for other architectures
• Send Swarm logs to stdout

1.2.6-rc1

API

• Switch from engine-api to docker/api/type and docker/clients
• Pass size parameter through on container inspect

Scheduler

• Precompile filter regular expression to reduce PU usage
• Remove setTCPUserTimeout to avoid TCP connection leak
• Fix network endpoints for rescheduling
• Give up leadership when manager shuts down

MISC

• Migrate documentation to https://github.com/docker/docker.github.io/tree/master/swarm
• Update Swarm CI to use go 1.7.1
• support GOARCH to be able to build for other architectures
• Send Swarm logs to stdout

v1.2.5

1.2.5 (2016-08-18)

Scheduler

• Fix container rescheduling with overlay network
• Fix scheduler detail log improper effect when container name is empty
• Check unique container name on create and rename for Mesos cluster

Health check

• Refresh container status on health_status events

Doc

• Fix install-w-machine.md using docker-machine --swarm feature

v1.2.4

1.2.4 (2016-07-28)

API

• New client interface in Swarm, to differentiate from Swarm mode in Docker 1.12
• Underlying HTTP client for API is created inside Swarm
• Update minimum Docker Engine version supported by Swarm to 1.8
• Additional error handling
• Code refactoring

Networking

• Fix concurrent map writes race condition
• Refresh single network when network event is emitted (performance improvement)
• Avoid network refresh when creating container (performance improvement)

Volumes

• Refresh single volume when volume event is emitted (performance improvement)
• Avoid volume refresh when creating container (performance improvement)

Events

• Support daemon events for Swarm

Test

• Fix leader election tests
• Fix rescheduling test

Mesos

• Fix double locking issue

Misc

• Handle systime difference between Swarm and Engines
• Add healthcheck information to CLI
• Fix engine_reconnect issue that led to reconnected engine being treated as new

v1.2.4-rc2

1.2.4-rc2 (2016-07-20)

API

• New client interface in Swarm, to differentiate from Swarm mode in Docker 1.12
• Update minimum Docker Engine version supported by Swarm to 1.8
• Additional error handling

Networking

• Fix concurrent map writes race condition
• Refresh single network when network event is emitted (performance improvement)
• Avoid network refresh when creating container (performance improvement)

Volumes

• Refresh single volume when volume event is emitted (performance improvement)
• Avoid volume refresh when creating container (performance improvement)

Events

• Support daemon events for Swarm

Test

• Fix rescheduling test

Mesos

• Fix double locking issue

Misc

• Handle systime difference between Swarm and Engines
• Add healthcheck information to CLI
• Fix engine_reconnect issue that led to reconnected engine being treated as new