centurion

#193 Permit specifying an ssh_socket_heartbeat for the SSH mechanism. This can alleviate stalled deploys in widely-threaded execution.

#190 Lower logging level to DEBUG for container inspect output during deploy. This prevents printing env vars in deployment logs, as those vars often contain secrets.

New in this release:

• #189: Allow setting the pid namespacing mode on containers.

New in this release:

#188 Environment variables may be sourced from lambdas, which are evaluated during deploy time. This permits calling functions to specify different environment variables per host.

SSH support is considered stable, as it has performed well for several months.

One new fix in this release: https://github.com/newrelic/centurion/pull/187 adds the unless-stopped restart policy.

This release adds beta SSH support for teams deploying to Docker hosts which do not expose their TCP sockets over the network.

Two fixs and one feature, for a balanced diet.

Fixed: #164 via #177 - containers in host networking mode may now define a public port to be used during healthchecks. The port mapping will be passed through to Docker, but it should be ignored by the server.

Fixed: #128 via #147 - manually set container names were not obeyed, now they are. If you have a name: set in your deploy tasks, please read the notes on #147 to avoid having duplicate containers present after a deploy, especially in host networking mode. You should comment out the name: directive until you're ready to manually intervene as containers may not be found during a stop operation.

Feature: #170 Container label support! You may now specify labels to be assigned to the containers Centurion creates. See the README for details.

This fixes a bad behavior in 1.8.8 where setting a custom stop_timeout caused Centurion to give up too soon waiting on the response from Docker. This would result in Excon timeouts during deploys.

The response timeout is now always 120 seconds longer than the Docker stop timeout.

See #171 and #175.

Fixed in this release:

• Custom stop_timeout settings could not exceed 120 seconds due to underlying Excon default. Note that docker would still have waited for the given time before killing the container, but Centurion would give up before that time had passed. Now Docker and Centurion should agree on how long to wait for containers to stop.
• Documentation corrections.

Minor bugfix release.

This is a minor release that includes a few readme/doc changes + ipc_mode addition to the dsl.

Changes: https://github.com/newrelic/centurion/compare/v1.8.5...v1.8.6

This release includes a couple of minor updates/additions and a cgroup bugfix.

• Avoid double health checks
• Added a repair task (redeploy to unhealthy hosts during rolling deploy)
• Minor log message cleanup
• After the most recent large Service class refactor, centurion no longer set Cgroup params for memory or CPU shares. This has now been fixed. NOTE: If you rely on this functionality - please docker inspect your running containers and verify that Memory is not 0 or max possible memory; redeploy the container if it's incorrect.

Changelog: https://github.com/newrelic/centurion/compare/v1.8.4...v1.8.5

This is a minor bugfix release:

• Added support for docker exec's
• Updated public_port_for() to utilize the correct port_binding structure (since latest refactor)
• Updated/fixed stop_containers() to correctly lookup containers by:
  • by name IF no public_ports are defined
  • Otherwise, lookup containers by first public_port

Changelog: https://github.com/newrelic/centurion/compare/v1.8.3...v1.8.4

Quick release that adds support for Docker 1.7 and 1.8 (and retains backwards compat with 1.5).

Changelog: https://github.com/newrelic/centurion/compare/v1.8.2...v1.8.3

This release fixes a bug in the rolling_deploy code which was no longer waiting on the set timeout intervals. Additionally output is fixed to report proper hostname strings rather than objects.

This fixes a bug with host-based networking deployments.

A lot of minor updates went into this release, and the code is now stable on the major internal refactor that was done by @kremso. The major additions are support for capabilities, host-based networking, and support for restarting containers without a redeploy.

Here is the changeset

There is one notable change in this release. The container naming will no longer default to using the Docker server's hostname inside the container. It will leave it as the container ID which is the Docker default. You can supply an optional lambda or Proc to name it as you see fit. The README contains full details.

Here is the changeset