operable / go-relay

Cog Relay written in Go supporting Docker and native command bundles

Relay

Dependencies

• Go v1.9+
• Docker v1.10.3+

Getting up and running

1. Clone go-relay to $GOPATH/src/github.com/operable/go-relay

   mkdir -p $GOPATH/src/github.com/operable
   git clone [email protected]:operable/go-relay.git $GOPATH/src/github.com/operable
   

2. Install preqrequisites

go get -u github.com/kardianos/govendor

3. Download deps and compile an executable

   make
   

4. Set environment variables and run relay.

   You'll need to have a docker machine running and have environment variables set for the docker client to connect to it. If you haven't already, run the following. (See more details at https://docs.docker.com/machine/get-started)

   docker-machine create --driver virtualbox default
   docker-machine start default
   eval $(docker-machine env default)`
   

   Then start relay:

   RELAY_DOCKER_USE_ENV=true _build/relay -file example_cog_relay.conf
   

Docker Images

Release images are available from the operable/relay repository on Docker Hub.

The latest code from the master branch is always available in the operable/relay-preview repository. The only tag in this repository is latest, and it "floats".

Building your own image can be done with make docker.

go-relay's Issues

Add support for Amazon's EC2 Container Registry (ECR) to Relay

From @christophermaier on December 30, 2016 19:25

Cog users wishing to access container images stored in Amazon's EC2 Container Registry from Relay will run into issues because a docker login yields a token that is effective for only 12 hours. With Relay's current implementation, you'd likely need to resort to things like restart the Relay process on a schedule in order for it to continue working.

Relay should have some smarts to contend with this use case. @kevsmith suggested exposing ECR support as a new engine within Relay.

Copied from original issue: operable/cog#1281

Segfault with managed dynamic configs

After upgrading the relay to ab7647a (the latest master, as of now), it started throwing segfaults:

Aug 13 14:00:32 foo relay[557]: time="2018-08-13T14:00:32+03:00" level=info msg="Using config file '/etc/relay.conf'."
Aug 13 14:00:32 foo relay[557]: time="2018-08-13T14:00:32+03:00" level=info msg="Relay 8b3dcfd8-41ec-44f7-a5ac-a3c1652fe9c1 is initializing."
Aug 13 14:00:32 foo relay[557]: time="2018-08-13T14:00:32+03:00" level=info msg="Started 16 request workers."
Aug 13 14:00:33 foo relay[557]: time="2018-08-13T14:00:33+03:00" level=info msg="Managed bundle dynamic configs enabled."
Aug 13 14:00:33 foo relay[557]: time="2018-08-13T14:00:33+03:00" level=info msg="Refreshing bundle dynamic configs every 5s."
Aug 13 14:00:35 foo relay[557]: panic: runtime error: invalid memory address or nil pointer dereference
Aug 13 14:00:35 foo relay[557]: [signal SIGSEGV: segmentation violation code=0x1 addr=0x28 pc=0x7404e8]
Aug 13 14:00:35 foo relay[557]: goroutine 1 [running]:
Aug 13 14:00:35 foo relay[557]: github.com/operable/go-relay/relay.(*DynamicConfigUpdater).refreshConfigs(0xc4203e0000)
Aug 13 14:00:35 foo relay[557]: /Users/agis/dev/go/src/github.com/operable/go-relay/relay/dynamic_config.go:104 +0xd8
Aug 13 14:00:35 foo relay[557]: github.com/operable/go-relay/relay.(*DynamicConfigUpdater).Run(0xc4203e0000, 0x24, 0xc420016009)
Aug 13 14:00:35 foo relay[557]: /Users/agis/dev/go/src/github.com/operable/go-relay/relay/dynamic_config.go:55 +0x19c
Aug 13 14:00:35 foo relay[557]: github.com/operable/go-relay/relay.(*cogRelay).handleBusEvents(0xc420010540, 0x87a600, 0xc420296d80, 0x0)
Aug 13 14:00:35 foo relay[557]: /Users/agis/dev/go/src/github.com/operable/go-relay/relay/relay.go:133 +0x5bc
Aug 13 14:00:35 foo relay[557]: github.com/operable/go-relay/relay.(*cogRelay).(github.com/operable/go-relay/relay.handleBusEvents)-fm(0x87a600, 0xc420296d80, 0x0)
Aug 13 14:00:35 foo relay[557]: /Users/agis/dev/go/src/github.com/operable/go-relay/relay/relay.go:79 +0x48
Aug 13 14:00:35 foo relay[557]: github.com/operable/go-relay/relay/bus.(*MQTTConnection).Connect(0xc420296d80, 0xc42025b920, 0x2e, 0xc420018010, 0x10, 0xc42001603f, 0x15, 0x75b, 0x0, 0x0, ...)
Aug 13 14:00:35 foo relay[557]: /Users/agis/dev/go/src/github.com/operable/go-relay/relay/bus/mqtt.go:51 +0x3b0
Aug 13 14:00:35 foo relay[557]: github.com/operable/go-relay/relay.(*cogRelay).Start(0xc420010540, 0x878240, 0xc420010540)
Aug 13 14:00:35 foo relay[557]: /Users/agis/dev/go/src/github.com/operable/go-relay/relay/relay.go:91 +0x4f9
Aug 13 14:00:35 foo relay[557]: main.main()
Aug 13 14:00:35 foo relay[557]: /Users/agis/dev/go/src/github.com/operable/go-relay/main.go:182 +0x214

After bisecting, it's apparent that 6a9d3ec caused this.

We do have the managed dynamic configuration mode enabled.

Upgrade Go version to v1.9

Go 1.9 is now stable release. The relay is still on 1.6.3-r0, and there's been a number of improvements in compiler and runtime since then.

Announcement here: https://blog.golang.org/go1.9

Additionally there's now an official package manager with https://github.com/golang/dep upgrading will allow us to use.

Change flag for custom relay config location to `-c, --config`

From @vanstee on May 4, 2016 21:23

Copied from original issue: operable/cog#628

Cannot pull circuit-driver when running against a private registry

Log output is

8/28/2017 4:56:31 PMtime="2017-08-28T20:56:31Z" level=debug msg="Retrieving operable/circuit-driver:0.13 from upstream Docker registry."
8/28/2017 4:56:31 PMtime="2017-08-28T20:56:31Z" level=error msg="Error ocurred pulling image operable/circuit-driver: Error response from daemon: Get https://registry-1.docker.io/v2/operable/circuit-driver/manifests/0.13: unauthorized: incorrect username or password."
8/28/2017 4:56:31 PMtime="2017-08-28T20:56:31Z" level=error msg="Error response from daemon: Get https://registry-1.docker.io/v2/operable/circuit-driver/manifests/0.13: unauthorized: incorrect username or password"

I have a relay set up to pull from GitLab's registry for container images and as such it has the RELAY_DOCKER_REGISTRY_EMAIL, RELAY_DOCKER_REGISTRY_HOST, RELAY_DOCKER_REGISTRY_PASSWORD, and RELAY_DOCKER_REGISTRY_USER set as environment variables. However, this appears to be preventing the relay from pulling from registry-1.docker.io as it's trying to use the username and password from the GitLab registry.

If another relay has launched on this host and the circuit-driver container image is cached, then it launches fine.

Environment variables set by Dockerfile don't get passed to the command

If a Dockerfile sets ENV vars they are not actually available to the running command. This doesn't seem like expected behavior, and makes testing very inconsistent (testing locally in a container acts differently than it does in cog).

Add Relay CLI tool to run commands as if you were talking to Cog

From @imbriaco on February 29, 2016 18:51

Cog commands are simple executables that are executed by Relay, with their arguments and options validated and populated in a specially designed environment. We should explore creating standalone utilities for executing these on the command line in the event that Cog is unavailable for any reason (chat provider outage, database failure, etc).

High level notes:

• Simple CLI tool converts command line arguments for use by Cog and populates the environment accordingly. e.g. cogexec mist:ec2-find --region=us-east-1.
• Other functions: list bundles, show help for a bundle. We have the necessary information available in the bundle configuration files.
• Pipelines are explicitly out of scope. Single command invocations only.
• No permission validation. Permission is assumed If users can login to the Relay host and have necessary system permissions to execute cogexec and access bundles, command_config, and related files.
• Any action performed by cogexec should be logged to the audit log.

Much more definition is necessary, but this captures the essence of the problem we're trying to solve.

Copied from original issue: operable/cog#321

Tests failing

Looks like we need to update the go version to fix this as 1.8 isn't supported anymore.
Will look into this when I have a chance.

relay smoke test

From @mpeck on November 22, 2016 21:7

Note: This will most likely be one of several issues for smoke testing communications between the various components of Cog.

Cog admins need an easy way to smoke test a Cog installation. Previously we were using Docker's healthcheck hook to provide some insight into Cog's status, but that proved to be a bit too broad to be useful. So instead we will provide a simple smoke test script to be installed alongside relay. This should give admins a bit more flexibility and provide a sanity check during installation and/or troubleshooting.

The smoke test script should be easily accessible via docker exec so we should copy it to somewhere in the path when building the image.

Checks

• the host running Cog is accessible
• the services endpoint in accessible

Done Criteria

• write a simple smoke test script
• update Dockerfile to install the script in the path (admins should be able to run the script with something like docker exec cog_smoke_test
• Add usage documentation to Cog book

Copied from original issue: operable/cog#1176

Relay suddenly begins refreshing commands way too fast

Currently my relay is set to refresh commands every 30 seconds, at some point it starts updating every 1-2 seconds which is much too fast and causes some high load.

This feels like a problem with the timer but seems simple enough that I'd be surprised if that were the case....
https://github.com/operable/go-relay/blob/master/relay/relay.go#L256