zoidberg-v0.6.1

• Migrate to dep for dependency management
• Test latest stable and master in travis
• Update alpine to 3.6 and go to 1.8.3

zoidberg-v0.6.0

From this release forward each port creates a separate app. This allows you to use different load balancers for different ports of your app. See README for more details about configuration. You can still use the old form of global configuration where whole apps belong to only one load balancer.

This release is backwards incompatible with previous zoidberg-tcp installations.

Before upgrading zoidberg managing zoidberg-tcp you have to add labels in the new form to make it work for both versions. After that you can upgrade zoidberg, then zoidberg-tcp, and then finally drop legacy labels.

This version also has slightly different CLI flags.

zoidberg-v0.5.4

• Fix concurrent map write on balancer update

zoidberg-v0.5.3

• Discard debug log for marathon

zoidberg-v0.5.2

• Explicitly register flags for finders
• Update vendored dependencies, fix incompatibilities
• Implement lazy state updates

zoidberg-v0.5.1

• Register mesos balancer finder as mesos, not marathon.
• Correct state.Versions comment.
• Add String() metthod to application.Server.
• Mention zoidberg-tcp in readme.

zoidberg-v0.5.0

This is another backwards-incompatible release in terms of CLI args. Zoidberg was heavily refactored inside, so things could break.

• Application and balancer finders are now separate and pluggable. You can discover balancers on Mesos and applications on Marathon. One command to rule them all.
• Metadata can be attached to applications to make better balancing decisions.
• Version is set to 1 by default.
• Travis now checks our sanity to some degree.

zoidberg v0.4.3

• Add explicit 5 second timeout for balancer and mesos communication

zoidberg v0.4.2

Prevent null instead of empty list for servers. This made zoidberg-nginx unhappy.

zoidberg v0.4.1

This release introduces support for static list of load balancers in marathon discoverer. This is useful when your balancers don't change their addresses and there is no need to launch them on Marathon.

This also enables easier multi-marathon deployments where each marathon gets zoidberg instance, but they all share the same set of load balancers.