- Fix missing HostConfig for rescheduled containers
- Fix TCP connections leak
- Support
docker run --net <node>/<network> ...
- Fix CORS issue in the API
- Fix regression with Docker Compose
- Performance improvements around networking with Docker engine 1.10
- Fix reschedule issue regarding events
- Implement engine refresh backoff strategy for failing nodes
- Add support for container rescheduling on node failure. (experimental)
- Use failureCount as a secondary health indicator
- Add swarm container create retry option
- Fixed the way soft affinities and handled
- Support private registry on docker run
- Expose error and last update time in docker info
- Sort images by Created
- Fix error when inspect on unhealthy node
- Prevent panic in filters when container has no name
- Add buildtime, kernelversion and experimental to API version
- Support docker update and new networking related flags in run & network create/connect/disconnect/ls
- Require
--all
on docker ps to display containers on unhealthy nodes - Retry on docker events EOF
- Add a random delay to avoid synchronized registration at swarm join
- Use engine connection error to fail engine fast
- Introduce pending state
- Rename slave to agent
- Upgrade tests to use mesos 0.25
- Code refactors
- Improve debug output
- Enable checkpoint failover in FrameworkInfo
- Fix timeout when pulling images
- Add timeout to refuse offers
- Fix double start issue
- Fix license grant
- Documentation update
- Use discovery from docker/docker
- Set labels for pending containers to fix scheduler failure
- Increase default TTL and heartbeat values to reduce traffic to discovery
- Fix 'ps -a' panic issue
- Fix network connect/disconnect for overlay network
- Fix connection leak on TLS connections
- Fix CLI hang on events command
- Fix newline issue with events
- Improve OPTIONS handler
- Fix image digest
- Fix memoryswappiness default value
- Enable profiling for HTTP in debug mode
- Provide options on swarm node update frequency
- Change offers timeout default to prevent other frameworks starvation
- Improve error output for bad swarm mesos user
- Fix connection failure when using Mesos with ZooKeeper
- Update to Go 1.5.2
- Documentation update
- Swarm is now pulling Images in parallel after the scheduling decision has been made, which mitigates the error happening occasionally with Docker pulls blocking the entire scheduler.
- The Node refresh loop process has been improved and does not yield to a panic when removing an Engine and trying to refresh the state at the same time.
- The refresh loop has been randomized to better handle huge scale scenarios (> 1000 nodes) where a refresh burst could occur and make the Manager unstable/fail.
- General improvements and fixes for the Mesos experimental backend for swarm.
- It is now possible to use the new networking features with Swarm through the
network
sub-system. Commands likedocker network create
,docker network attach
anddocker network ls
are now available through swarm.
- You can now use the docker volume plugin subsystem available with
docker volume
.
- You can now specify the
--replication-ttl
flag to control how long it takes for Replicas to be notified of the Primary failure and take over the lead.
- This is now possible to use TLS with discovery for
consul
andetcd
using the new--discovery-opt
flag.
- Reschedule with soft affinity when image cannot be pulled
- Replace the store pkg by libKV
- Fixes about consul/etcd and zookeeper
- Fix docker push name matching
- Fix docker exec status code
- Fix docker pull status code
- Improve docker info with docker client 1.7.x
- Add SystemTime, http_proxy, https_proxy and no_proxy to docker info
- Task creation timeout configurable
- Fix issue with hostname in library image
- Use 'docker_port' attribute if available
- Add support for random ports
- Add doc on leader election / high availability of swarm manager
- Lots of typos/improvements to the doc
- Switch to golang 1.4