#codefresh (2019-04)

So it turned out I was using ‘global variables’ and build variables wrong

I thought this would mean when I run the build pipeline my Dockerfile’s ARG PORT would get a value of 3060

Turns out it doesn’t

So I found this: https://codefresh.io/docs/docs/yaml-examples/examples/build-an-image-with-build-arguments/

and that’s fine.. happy to define it in a codefresh.yml file

but then I realised how do I do multi-branch / stage CI… surely I’ll need multiple codefresh.yml because one example argument is APP_ENV

I was going to have a pipeline for each stage inside the codefresh UI which runs off the same codefresh.yml and simply has the build args in the UI

Does any of this sound like I’ve done / misunderstood it incorrectly, as having multiple codefresh.yml doesn’t feel DRY.. especially to only change a few build args. Thanks

Ah.. unless it should be ENV instead of ARG idk

how many variables are you setting at build time? I usually just use environment variables that are set at run time instead of build time. I add the environment variables through the pod definition. This allows my Dockerfile to be pretty generic

So far 2. Vant see it breaching 5

I see your screenshot that your are setting aws secrets and the port for your container. I don’t think the port matters on your Docker container, so I am assuming its more about your aws secrets. I never was a fan of keeping keys lying around, and I liked having everything controlled through IAM. May I suggest taking a look at https://github.com/uswitch/kiam ? That’s what we use to make pods have specific IAM permissions, and it works great.

Thanks Mark. I’ve actually removed it now. It was when I was trying put ECS

The two args I have are: application port and application env file.

Because I then do: COPY ${APP_ENV} .env

How would you go about checking if .env is a directory or file inside a docker image inside a pipeline?

You could have a freestyle step running a bash IF -f or -d; you could have some sort of condition in Dockerfile maybe?

And also to add geodesic to CI, good to just use https://github.com/cloudposse/prod.cloudposse.co/blob/master/codefresh.yml ?

Seems to be only pushing to codefresh registry which works for me. Unsure why it defines an old build harness?

Erik, where’s an example of you executing a command in a geodesic module from a freestyle step in codefresh?

https://github.com/cloudposse/testing.cloudposse.co/pull/22/files

We should move that “use_codefresh” direnv function into geodesic

thanks

I get this funny bash script actually

@oscarsullivan_old that’s the default behavior of geodesic

that’s why you can run

docker run myco/myinfra | bash

to install geodesic

to avoid that, do this:

what did I miss there Erik?

https://github.com/cloudposse/testing.cloudposse.co/pull/22/files#diff-fee4a8e1fcec1aac118c7e23c3e707ebR25

because command not cmd?

i am not sure about command vs cmd

https://codefresh.io/docs/docs/codefresh-yaml/steps/freestyle/

looks like cmd is canonical

ah I went from this

CollectAllMyDeps:
  title: Install dependencies
  image: python:3.6.4-alpine3.6
  commands:
    - pip install .

Ok will try to make it closer looking to what you’ve got

oh, so yea, cmd is the arg passe to the entrypoint

could be the difference

vs commands is run after the entrypoint

ahhh

you can try this:

that makes sense

although that confuses me again

no

hrm… maybe

seems dull I can’t just have commands: ansible --version

and have to split it up

nono

I’ll give these new combos a go

in that example, you are passing --version to the entrypoint which is bash

so you should get the version of bash back

ooh

you could alternatively change the entrypoint

study up on ENTRYPOINT vs CMD in docker

then see what we’re doing in the Dockerfile for geodesic

study […] Thanks

Must admit I don’t really get the differences

Seen both in action but no further

yea, the subtle nuances are often misunderstood by Dockerfile authors

and you see them misused and abused

I liken it to the system call

int execve(const char *filename, char *const argv[],
                  char *const envp[]);

ENTRYPOINT ~ filename

CMD ~ argv

(this is my mental model, not the docker explanation)

lol Erik it was because I had command not commands

  DeployingDockerImage:
    title: Deploying Docker Image with Ansible
    image: r.cfcr.io/xxx/xxx/${{MODULE}}.xxxx.co.uk:master
    commands:
      - ansible --version 
      # ansible-playbook pod.yml -i inventory/${{inventory}}.yml

Output:

Status: Downloaded newer image for r.cfcr.io/xxx/xxx/sandbox.xxx.co.uk:master                               
ansible 2.7.9                                                                                                              
  config file = None                                                                                                       
  configured module search path = [u'/conf/.ansible/plugins/modules', u'/usr/share/ansible/plugins/modules']               
  ansible python module location = /usr/lib/python2.7/site-packages/ansible                                                
  executable location = /usr/bin/ansible                                                                                   
  python version = 2.7.15 (default, Jan 24 2019, 16:32:39) [GCC 8.2.0]                                                     
Reading environment variable exporting file contents.                                                                      
Successfully ran freestyle step: Deploying Docker Image with Ansible                              

Now I gotta get my private Role + Playbook accessible

Thinking either have them in conf, have some kind of wget situation and publicise them (not ideal), try and do an import like with terraform modules?????

Or my geodesic module has a git clone.. but then where do I store the deploy key hmmmm

Have come up with this solution

You could also clone as a step in the codefresh pipeline

that would be accessible from inside of geodesic

the benefit of doing it that way is that you can leverage the git integrations already available in codefresh without putting a secret (e.g. /conf/ssh/deployment-repo-key into git)

Oh wow

yeh you’re right

that’s perfect actually. thanks

Webhooks issue Apr 3, 09:27 UTC Resolved - The incident been resolved Apr 3, 09:24 UTC Investigating - We are currently investigating an issue where git webhooks don’t trigger pipelines. Running builds manually works as expected

Codefresh availability issue Apr 3, 12:29 UTC Investigating - We are currently investigating a problem affecting Codefresh availability

Codefresh availability issue Apr 3, 12:46 UTC Monitoring - Codefresh site is up, we’re still monitoring the system Apr 3, 12:29 UTC Investigating - We are currently investigating a problem affecting Codefresh availability

Codefresh availability issue Apr 3, 13:34 UTC Update - Google just updated us on an ongoing incident, we’re monitoring the issue with google team Apr 3, 12:46 UTC Monitoring - Codefresh site is up, we’re still monitoring the system Apr 3, 12:29 UTC Investigating - We are currently investigating a problem affecting Codefresh availability

Codefresh availability issue Apr 3, 14:23 UTC Resolved - This incident has been resolved. Apr 3, 14:22 UTC Update - We are continuing to monitor for any further issues. Apr 3, 13:34 UTC Update - Google just updated us on an ongoing incident, we’re monitoring the issue with google team Apr 3, 12:46 UTC Monitoring - Codefresh site is up, we’re still monitoring the system Apr 3, 12:29 UTC Investigating - We are currently investigating a problem affecting Codefresh availability

what format do you use for your Secrets Manager in #aws to connect to the Codefresh private repo? Can anyone please share an example

Status: Image is up to date for cloudposse/build-harness:0.18.0                                                            
make: *** No rule to make target 'codefresh/notify/slack/deploy'.  Stop.                                                   
[SYSTEM] Error: Failed to run freestyle step: Send notification to Slack channel; caused by NonZeroExitCodeError: Container
 for step title: Send notification to Slack channel, step type: freestyle, operation: Freestyle step failed with exit code:
 2    

  SendSlackDeployNotification:
    title: Send notification to Slack channel
    stage: "deploy"
    image: cloudposse/build-harness:${{BUILD_HARNESS_VERSION}}
    commands:
      - make codefresh/notify/slack/deploy

Anyone aware of why this would be happening? LGTM according to build-harness readme and example here: https://github.com/cloudposse/example-app/blob/29f91a718522e4a702d77d172a41ed1f779d42fe/codefresh/pull-request.yaml#L107

This also seems to follow the same pattern https://docs.cloudposse.com/release-engineering/cicd-process/build-charts/#examples

likewise make init fails

@Igor Rodionov

So have tried the Codefresh commands of:

    ...
    commands:
      - make init

and as above

do you have declated env vars https://github.com/cloudposse/example-app/blob/29f91a718522e4a702d77d172a41ed1f779d42fe/codefresh/pull-request.yaml#L24

?

@oscarsullivan_old ^

check this env vars

https://github.com/cloudposse/example-app/blob/29f91a718522e4a702d77d172a41ed1f779d42fe/codefresh/pull-request.yaml#L24

https://github.com/cloudposse/build-harness/blob/master/modules/slack/Makefile#L17

https://github.com/cloudposse/build-harness/blob/master/modules/slack/Makefile#L17

To debug it would be useful to see env vars and codefresh.yaml

@oscarsullivan_old I just setup a pipeline using the codefresh/notify/slack/deploy/webapp target using BUILD_HARNESS_VERSION=0.18.0

I would check your environment variables and make sure they are set correctly, I was getting similar errors, and it was because of env variables not being set

Thanks, I’ll double check both these points (version and undeclared variables)

it is most likely the declaration of env vars

I did no special setup to use the build-harness tmeplates

does anyone have suggestions for the strategy I should use for deploying to production from codefresh? I feel unsafe giving the ClusterRole codefresh-role permission to deploy to the whole cluster

Or is that the only way possible?

I think I use this

but not using K8s rn

We’ve added support for DocumentDB (MongoDB) to our Codefresh Enterprise terraform module: https://github.com/cloudposse/terraform-aws-codefresh-backing-services

This makes everything that is absolutely essential to running Codefresh Enterprise (onprem) a fully managed service by AWS.

Awesome! I made this little example app a little more deployable… Still working out the kinks. https://github.com/codefresh-contrib/example-voting-app

Requires a few IPs right now and only works with Cloud LB not sure how to make this more portable using Istio or NGINX as both would require a DNS integration using the LB as is while requiring 2 IPs per running Helm Release is a bit much but this is also meant to demo a few things then toss after playing.

Bitbucket access issue Apr 11, 11:15 UTC Investigating - We are currently investigating a problem affecting bitbucket integrations

Build services disruption Apr 11, 13:11 UTC Investigating - We are currently investigating this issue

Bitbucket access issue Apr 11, 14:43 UTC Resolved - The incident has been resolved Apr 11, 11:15 UTC Investigating - We are currently investigating a problem affecting bitbucket integrations

Build services disruption Apr 11, 14:43 UTC Resolved - The incident has been resolved Apr 11, 13:11 UTC Investigating - We are currently investigating this issue

https://cloudposse.com/slides/gitops-with-terraform-on-codefresh-webinar/

Here are the slides from the webinar today.

Video will be posted as soon as it’s available.

Thank you

Here’s what a PR looks like: https://github.com/cloudposse/testing.cloudposse.co/pull/75

Here’s the configuration: https://github.com/cloudposse/testing.cloudposse.co/tree/master/codefresh/terraform

Helm repositories issue Apr 11, 19:04 UTC Resolved - A fix has been deployed to production and confirmed to have fixed the issue. Helm repositories are fully operational. Apr 11, 18:49 UTC Monitoring - The issue has been identified, and the corresponding fix was applied. Helm repositories are accessible again, We’re closely monitoring the platform to ensure everything is working as expected. Apr 11, 18:47 UTC Identified - Currently Helm repositories provided by Codefresh are having issues. This issue is under…

@Erik Osterman (Cloud Posse) Do you happen to have an example of running https://github.com/cloudposse/github-commenter as a step in Codefresh pipeline using a Docker image? I’ve seen in other project snippets of codefresh.yml formatted examples but nothing in this project.

Came up with this really quickly.

  GitHubCommenter:
    title: Add GitHub Comment
    image: cloudposse/github-commenter:latest
    environment:
      - GITHUB_TOKEN=${{GITHUB_TOKEN}} #Must be created see link for info below.
      - GITHUB_OWNER=${{CF_REPO_OWNER}}
      - GITHUB_REPO=${{CF_REPO_NAME}}
      - GITHUB_COMMENT_TYPE=pr
      - GITHUB_PR_ISSUE_NUMBER=${{CF_PULL_REQUEST_NUMBER}}
      - GITHUB_COMMENT="" #Your custom comment goes here

this is the latest example that Erik did for the presentation https://github.com/cloudposse/testing.cloudposse.co/blob/osterman-patch-1/codefresh/terraform/pipeline.yml

the PR looks like this https://github.com/cloudposse/testing.cloudposse.co/pull/75

Thanks!

uses this template file for github-commenter https://github.com/cloudposse/testing.cloudposse.co/blob/osterman-patch-1/codefresh/terraform/pipeline.yml#L59 (update it for your needs)

https://github.com/cloudposse/testing.cloudposse.co/blob/osterman-patch-1/codefresh/terraform/comment.txt.gotmpl

https://m.youtube.com/watch?v=AUOV0XMR_9Q

Is there a how-to on helmfile in Codefresh like an example codefresh.yml with an explanation on the steps?

Not that well documented

https://github.com/cloudposse/example-app/tree/master/deploy

The “ctl” command can be ignored. It just shows how to call Helmfile

In this example, we support both blue/green and rolling

Also, this example shows how we use the monochart

I’d be happy to jump on a call and walk you through it

How are you guys running stuff like integration tests and system tests? CF can’t directly access internal services that are deployed, so are you guys using some kind of proxy container in the cluster to run these kinds of tests?

https://github.com/codefresh-io/venona

This allows you to run it all in your k8s cluster

we’ve used this for things like connecting to artifactory or using consul

Yup this runner allows you to talk to the internal pods of the Kubernetes cluster using the container ips if needed.

this will just show up as a different run time environment on your pipeline?

Add the runner to the cluster and then select the runner for the pipeline and everything happens on your cluster.

Yes

https://codefresh.io/docs/docs/enterprise/behind-the-firewall/

Is there a way to split up what runs on CF runtime env and what runs on our infra? I assume that venona still reports everything up to CF UI, right?

Yes all comes back to UI

Each pipeline has a runtime setting. You can use our CLI to call another pipeline from your original as a step in the original and have that child pipeline run on your infra.

Has anyone been using cf_export stating the name of an existing environment variable? I have tried to use it, but the variable output is getting altered. FreeStye Bash Code

 - FEATURE=$(echo ${{CF_BRANCH}} | cut -c 9- )
      - echo $FEATURE
      - cf_export $FEATURE

Output

TI-595                                                                                                                                                  
Exporting TI-595=-595

I am seeing the cf_export command parse the static environment variable value vs. storing the defined value with the assignment cf_export $FEATURE.

@Michael Kolb here’s a pipeline step example to setup ENV vars

it exports Codefresh env vars and also those defined in Codefresh UI (Global config section and ENV vars section in each pipeline)

you mean the example does not use the ${{...}} syntax?

So, in the documentation it says that you can state an existing variable state the name of an existing environment variable (like EXISTING_VAR) https://codefresh.io/docs/docs/codefresh-yaml/variables/#using-cf_export-command

However, this hasn’t stored that variable assignment

and the displayed use case does not show assignment in the code

is the documentation referencing a pipeline variable defined in the pipeline vs. in the pipeline yaml?

Yes pipeline variables are in the pipeline SPEC file. If you need variables for the pipeline to be set during pipeline execution then you’d use cf_export to have them pop out to pipeline.

The existing variable section is mentioning step-2 in the YAML above.

You do not need to specify a variable name before the cf_export

thanks, the documentation page could use more examples of cf_export variable assignment like the example that you provided - cf_export GITHUB_OWNER=${{CF_REPO_OWNER}}

they display assignment as static references

and the reference EXISTING_VAR is misleading if this is pulling from the pipeline variable definition vs. an existing variable

documentation clarification would help, since I have had to run several builds trying to figure out how the cf_export process handles environment variables

is there any documentation on the codefresh/kube-helm image that provides connection to Kubernetes & Tiller?

Are you talking about the image used in our UI steps https://github.com/codefresh-contrib/images/tree/master/kube-helm or Helm step https://github.com/codefresh-contrib/cfstep-helm

the docker hub release

i found the command that i need kubectl config use-context <<cluster name>>

the container had all the tools installed. Unfortunately, there was not documentation on the docker hub site.

Hi.. anyone got experience using submodules and codefresh?

So it gets my git token, updates the submodules in /codefresh/volume/my_app

then does an ls pf those dirs to show me it has cloned (it has)

and then in the docker build step I have: COPY models/* models/

but the models/ dir is empty in the container..

So I can only assume that /codefresh/volume/my_app is not the path that my_app is being cloned into during init and therefore not the context of Docker when building

I would recommend trying to set the working_directory: ${{main_clone}} for the step to ensure you’re in the clone repository directory.

Docker incident: Docker Hub elevated errors Apr 25, 17:33 UTC Investigating - Docker has reported an incident affecting Docker Hub Web and Docker Hub Automated Builds components. It shouldn’t affect normal pulling and pushing operations. We will keep monitoring this.

More information here: https://status.docker.com/

Docker incident: Docker Hub elevated errors Apr 25, 18:10 UTC Resolved - Docker has reported this incident as Resolved. “[Resolved] The issue has been resolved. The DB is backup again.”

More information here: https://status.docker.com/pages/incident/533c6539221ae15e3f000031/5cc1ed78790d0e1ca1c8fcd4 Apr 25, 17:33 UTC Investigating - Docker has reported an incident affecting Docker Hub Web and Docker Hub Automated Builds components. It shouldn’t affect normal pulling and pushing operations. We will keep monitoring this.

More…

Does Codefresh have API that accepts commands to start an existing pipeline? I have an integration test that will take hours and would like the option to start another pipeline once the integration tests complete.

https://codefresh-io.github.io/cli/pipelines/run-pipeline/

Usage from a step in pipeline.

https://github.com/codefresh-contrib/example-voting-app/blob/master/.codefresh/codefresh-cd.yml#L72-L76