v1.2.0

1.2.0 (2016-04-13)

Scheduler

• Move rescheduling out of experimental
• Differentiate constraint errors from affinity errors
• Printing unsatisfiable constraints for container scheduling failure
• Enable rescheduling on master manager to prevent replica managers from rescheduling containers
• Output error when starting a rescheduled container fails, and when removing container fails at node recovery
• Validate cluster swarm.overcommit option

API

• Introduce engine-api client to Swarm
• Implement 'info' and 'version' with engine-api
• Use apiClient for some volume, network, image operations
• Print engine version in Info
• Fix swarm api response status code
• Support ps node filter
• Fix HostConfig for /start endpoint
• Print container 'created' state at ps
• Update dockerclient to get labels on volumes, networks, images
• Support private images, labels and other new flags in docker build
• Select apiClient version according to node docker version

Node management

• Prevent crash on channel double close
• Manager retries EventMonitoring on failure.
• Docker engine updates hostname/domainname
• Force inspect for containers in Restarting state.
• Increase max thread count to 50k to accommodate large cluster or heavy workload
• Force to validate min and max refresh interval to be positive
• Skip unstable tests from Docker bug 14203
• Fix race condition between node removal from discovery and scheduler
• Fix data race with node failureCount
• Display warning message if an engine has labels with "node=xxx"

Discovery

• Remove parameter which is not used in createDiscovery
• Fix Consul leader election failure on multi-server

Mesos integration

• Support rescind offer in swarm
• Update mesos tests

Misc

• Update golang version to 1.5.4
• Skip redundant endpoints in "network inspect"
• Validate duration flags:--delay, --timeout, --replication-ttl
• Fix image matching via id
• Make port 0 invalid as listening port
• Improve volume inspect test
• Add read lock for eventsHandler when only it is necessary