Skip to content

Latest commit

 

History

History
222 lines (167 loc) · 13.6 KB

DO288.md

File metadata and controls

222 lines (167 loc) · 13.6 KB
Raw

Objectives DO288/EX288

Exam DO288 objectives:

  • Troubleshoot application deployment issues
  • Create custom image streams
  • Use OpenShift secrets and ConfigMaps to inject configuration data into applications
  • Customize existing S2I builder images
  • Implement hooks and triggers
  • Work with multi-container templates
  • Implement application health monitoring

Deploying an app

oc new-app

Option Description
--image-stream[-i] image stream to be used
--strategy docker or source
--code URL to Git repo that will be used as input for S2I
--docker-image URL to a container

Examples:

# OS inspects the code to determine which image to use
oc new-app http://gitserver.example.com/mygitrepo

# use a PHP S2I builder image (both commands are the same)
oc new-app php~http://gitserver.example.com/mygitrepo
oc new-app -i php http://gitserver.example.com/mygitrepo
# again, same with tag
oc new-app php:7.0~http://gitserver.example.com/mygitrepo
oc new-app -i php:7.0 http://gitserver.example.com/mygitrepo

# command detects the container image URL as the input argument
oc new-app registry.example.com/mycontainerimage
# to avoid detection problems, you can force the type of URL (code vs Docker image)
oc new-app --code http://gitserver.example.com/mygitrepo
oc new-app --docker-image registry.example.com/mycontainerimage
# to force use of Dockerfile instead of S2I
oc new-app --strategy docker http://gitserver.example.com/mydockerfileproject
# to force use of S2I instead of Dockerfile
oc new-app --strategy source http://gitserver.example.com/user/mygitrepo
# other options, such as --image-stream and --code, can be used in the same command with --strategy

oc new-app command adds the following resources to the current project:

  • build configuration
  • image stream
  • deployment configuration
  • service for all ports exposed by the application container image
  • app label based on the short name of the Git repo or --name
# following command deletes all resources created by the previous oc new-app command
oc delete all -l app=test
# if you want to inspect the resource definitions without creating the resources in the current project, you use the -o option
oc new-app -o json registry.example.com/mycontainerimage

Image Streams and Tags

  • Having container image metadata in an IS allows to perform operations based on this data instead of going to a registry server every time.
  • Allows using either notification or pooling strategies to react to image content updates.
  • IS resource can define multiple IS tags, tags can point to a different container image tag or even to a different container image name (e.g. the ruby IS from the openshift project defines the following IS tags: ruby:2.0, ruby:2.2 both referring to different images in different registries).
# easiest way to create an image stream is by using the oc import-image command with the --confirm option
oc import-image myis --confirm --from registry.acme.example.com:5000/acme/awesome --insecure
# create new app using Dockerfile in Git repo using base image in an insecure registry
oc new-app --name echo --insecure-registry http://services.lab.example.com/rhel7-echo

# build configuration output is a container image sent to the internal registry through an image stream
oc describe bc echo | grep "Output to:"
# Output to: ImageStreamTag echo:latest

# deployment configuration has a trigger in the image stream. If the image stream changes, a new deployment is performed
oc describe dc echo | grep "Triggers:"

# change the app, rebuild the application and verify that OpenShift deploys the new container image
oc start-build echo
# show the history of image streams
oc describe is echo

Managing Applications with the Web Console

  • The logs from a deployment are not stored by OpenShift unless there were errors during deployment.
  • Default OpenShift router implements session affinity (if you have multiple pods and you go to the router's URL and keep refreshing, you will be routed to the same pod every time).
  • If the application source code repository is not configured to use OpenShift web hooks, you need to use either the web console or the CLI to trigger new builds, after pushing updates to the application source code.
  • The details page for a deployment also provides the Deploy button, which performs the same function as the oc rollout latest command. You need to use the CLI to perform other oc rollout operations.
  • Deleting a project deletes all resources inside the project.

Managing Applications with the CLI

  • The default configuration for an OpenShift cluster allows any user to create new projects.

  • A user is automatically a project administrator for any project he creates.

  • Access levels:

Type Permissions
Cluster admin can manage projects, add nodes, create persistent volumes, assign project quotas, and perform other cluster-wide administration tasks.
Project admin can manage resources inside a project, assign resource limits, and grant other users permission to view and manage resources inside the project.
Developer can manage a subset of a project's resources. The subset includes resources required to build and deploy applications, such as build and deployment configurations, persistent volume claims, services, secrets, and routes. They cannot grant to other users any permission over these resources, and cannot manage most project-level resources such as resource limits.
  • Search for a template that uses PHP and MySQL: oc get templates -n openshift | grep php | grep mysql.
  • When you create a new template (oc create -f <name.json>) not in openshift namespace (-n openshift), it will only be available in the project where you created it.

Troubleshooting Builds, Deployments, and Pods Using the CLI

  • Differences between oc describe and oc get -o:
    • The oc describe command can follow relationships between resources (e.g. describing a build configuration shows information about the most recent build).
    • The oc get -o command shows information only about the requested resource (e.g. build configuration does not show information about the recent builds).
  • oc edit = oc get -o + vim + oc apply.
  • The logs from a build configuration (oc logs bc/myapp) are the logs from the latest build, whether it was successful or not, but you can use oc logs build/myapp-X to see the logs from the Xth build.
  • oc logs myapp-build-X to see the logs of the build pod created to perform the Xth build.
  • to retrieve the Apache HTTP Server error logs stored inside an application pod called frontend, use: oc cp frontend-1-abcde:/var/log/httpd/error_log /tmp/frontend-server.log, unlike the UNIX cp command, the oc cp command cannot copy a source file to a destination folder, only folder -> folder and file -> file.
  • oc cp command requires the application container image to provide the tar command. If not, the command will fail.
  • oc rsync command synchronizes local folders with remote folders from a running container. It uses the local rsync command to reduce bandwidth usage, but does not require the rsync or ssh commands to be available in the container image.
  • oc rsh command creates a remote shell to execute commands inside the container. It uses the OpenShift master API to create a secure tunnel to the remote pod, but does not use the ssh or the rsh UNIX commands.
  • Start an interactive shell session (-t) inside the container: oc rsh -t frontend-1-abcde.
  • Copy the SQL script to the database pod example: oc cp ./DO288/labs/build-template/quote.sql quotesdb-1-hh2g9:/tmp/quote.sql.
  • Run the SQL script inside the database pod example: oc rsh -t quotesdb-1-hh2g9 and then when inside: mysql -u$MYSQL_USER -p$MYSQL_PASSWORD $MYSQL_DATABASE < /tmp/quote.sql.

Troubleshooting three-tier app (HTTP/1.1 500 Internal Server Error when curl the router URL):

  1. Verify that the database service found the correct database pod: oc describe svc db | grep Endpoints and oc describe pod db-1-abcde | grep IP. Does it show the same IP?
  2. Verify the database pod login credentials: oc describe pod db-1-abcde | grep -A 4 Environment. e.g. MYSQL_USER, MYSQL_PASSWORD etc.
  3. Verify the database connection parameters in the application pod: oc describe pod api-1-abcde | grep -A 5 Environment. e.g. DATABASE_SERVICE_NAME, DATABASE_NAME etc.
  4. Verify that the application pod can reach the database pod: oc rsh api-1-abcde bash -c 'curl $DATABASE_SERVICE_NAME:3306'.
  5. Review application logs: oc logs api-1-abcde.

Build and Deployment Environment Variables

  • Add the -e option to the oc new-app command to provide values for environment variables. These values are stored in the deployment configuration and added to all pods created by a deployment.
  • For a builder pod, you define environment variables using the --build-env option of the oc new-app command.
  • Deployment configuration (dc) stores environment variables for application pods, while a build configuration (bc) stores environment variables for builder pods.

Designing containerized apps for Openshift

  • Preferred way to build images is S2I.
  • OpenShift provides the s2i command-line tool that helps you bootstrap the build environment for creating custom S2I builder images. It is available in the source-to-image package from the Red Hat Software Collections RHSCL Yum repositories (rhel-server-rhscl-7-eus-rpms).
  • Red Hat recommends to run the image as a non-root user for security reasons.
  • OpenShift, by default, does not honor the USER instruction set by the container image. For security reasons, a random userid other than the root userid (0) is used by OpenShift to run containers.

Building Container Images with Advanced Dockerfile Instructions

  • Each RUN in Dockerfile creates a new layer. It is recommended to use && on each line to create one long command and thus one layer.
  • Openshift uses lots of LABEL instructions (io.openshift.tags, io.k8s.description, io.openshift.expose-services).
  • When building images for Openshift, the namespace should be set to io.openshift.
  • WORKDIR should be absolute path.
  • It is a good practice to use ENV instruction(s) to define file and folder path(s) instead of hard-coding it. It is recommended to use one ENV and \ on each line to only create one layer.
  • USER sets the user name (or UID) and optionally the user group (or GID) to use when running the image and for any RUN, CMD, and ENTRYPOINT instructions that follow it in the Dockerfile.
  • VOLUME - mountpoint inside the container, used for persistent data, content is preserved even if the container is restarted or moved to a different node.
  • ONBUILD instruction registers triggers in the container image, declare instructions to be executed only when building a child image. Useful when you have image that everyone in the organization uses and build child images from it. The parent Dockerfile has ONBUILD instructions in this case and when the build process of the child image starts, it triggers the execution of the ONBUILD instructions in the parent image.

Adapting Dockerfiles for Openshift

  • Directories and files that are read from or written to by processes in the container should be owned by the root group and have group read or group write permission (RUN chgrp -R 0 <directory> && chmod -R g=u <directory>).
  • Files that are executable should have group execute permissions.
  • The processes running in the container must not listen on privileged ports (that is, ports below 1024), because they are not running as privileged users.
  • Example of the above:
# Use the do288/httpd-parent image as base
FROM registry.lab.example.com:5000/do288/httpd-parent

# Change the port to 8080
EXPOSE 8080

# Override the LABEL from parent
LABEL io.openshift.expose-services="8080:http"

# Change web server port to 8080
RUN sed -i "s/listen 80/listen 8080/g" /etc/nginx.conf

# Permission to allow container to run on Openshift
RUN chgrp -R 0 /var/opt/rh/rh-nginx18 && chmod -R g=u /var/opt/rh/rh-nginx18

# Run as a non-privileged user, Red Hat recommends this
USER 1001

Running Containers as root Using Security Context Constraints (SCC)

  • All containers created by OpenShift use the SCC named restricted by default, which ignores the userid set by the container image, and assigns a random userid to containers. You need anyuid SCC to allow containers run as root.
  • All pods from a project run under a default service account, unless the pod, or its deployment configuration, is configured otherwise.
  • Allow containers to run as the root user in an OpenShift project (not recommended):
oc create serviceaccount <myserviceaccount>
oc patch dc/demo-app --patch '{"spec":{"template":{"spec":{"serviceAccountName": "<myserviceaccount>"}}}}'
   OR oc edic dc/demo-app and add serviceAccountName attribute to the spec>template>spec section (below terminationGracePeriodSeconds)
oc adm policy add-scc-to-user anyuid -z <myserviceaccount>

Injecting Configuration Data into an Application

  • The recommended approach for containerized applications is to decouple the static application binaries from the dynamic configuration data and to externalize the configuration. This ensures the portability of applications across many environments.
  • Secrets and configuration maps must be created before creating the pods that depend on them.
oc create configmap config_map_name \
--from-literal key1=value1 \
--from-literal key2=value2

Last updated: Mon Feb 04 17:00:00 AEDT 2019 (notes from DO288-OCP3.6-en-1-20180130-ROLE.pdf, Progress 23.9% = page 95 of 394)