229 lines
7.8 KiB
Markdown
229 lines
7.8 KiB
Markdown
---
|
|
stage: Monitor
|
|
group: Observability
|
|
info: To determine the technical writer assigned to the Stage/Group associated with this page, see https://about.gitlab.com/handbook/product/ux/technical-writing/#assignments
|
|
---
|
|
|
|
# GitLab Observability Quickstart
|
|
|
|
You can try GitLab Observability by [cloning or forking the project](https://gitlab.com/gitlab-org/opstrace/opstrace.git) and creating a local installation.
|
|
|
|
## Prerequisites and dependencies
|
|
|
|
To install GitLab Observability Platform (GOP), install and configure the following third-party dependencies. You can do this manually, or [automatically by using asdf](#install-dependencies-using-asdf):
|
|
|
|
- [kind](https://kind.sigs.k8s.io/docs/user/quick-start/#installation) for creating a local Kubernetes cluster.
|
|
- [Docker](https://docs.docker.com/install)
|
|
- [Docker Compose](https://docs.docker.com/compose/compose-v2/) is now part of the `docker` distribution.
|
|
- [kubectl](https://kubernetes.io/docs/tasks/tools/#kubectl) for interacting with GitLab Observability.
|
|
- [Telepresence](https://www.telepresence.io/) allows you to code and test microservices locally against a remote Kubernetes cluster.
|
|
- [jq](https://stedolan.github.io/jq/download/) for some Makefile utilities.
|
|
- [Go 1.19](https://go.dev/doc/install).
|
|
|
|
The current versions of these dependencies are pinned in the `.tool-versions` file in the project.
|
|
|
|
You can run the following commands to check the availability and versions of these dependencies on your machine:
|
|
|
|
```shell
|
|
kind --version
|
|
docker --version
|
|
kubectl version
|
|
telepresence version
|
|
jq --version
|
|
go version
|
|
```
|
|
|
|
### Run GOP on macOS
|
|
|
|
If you're running GOP on macOS, ensure you have enough resources dedicated to Docker Desktop. The recommended minimum is:
|
|
|
|
- CPUs: 4+
|
|
- Memory: 8 GB+
|
|
- Swap: 1 GB+
|
|
|
|
It's possible to run GOP with fewer resources, but this specification works.
|
|
|
|
### Install dependencies using asdf
|
|
|
|
If you install dependencies using [`asdf`](https://asdf-vm.com/#/core-manage-asdf), GOP manages them for you automatically.
|
|
|
|
1. If you have not already done so, clone the `opstrace` repository into your preferred location:
|
|
|
|
```shell
|
|
git clone https://gitlab.com/gitlab-org/opstrace/opstrace.git
|
|
```
|
|
|
|
1. Change into the project directory:
|
|
|
|
```shell
|
|
cd opstrace
|
|
```
|
|
|
|
1. Optional. If you need to install `asdf`, run:
|
|
|
|
```shell
|
|
make install-asdf
|
|
```
|
|
|
|
1. Install dependencies using `asdf`:
|
|
|
|
```shell
|
|
make bootstrap
|
|
```
|
|
|
|
## Step 1: Create a local Kubernetes cluster with kind
|
|
|
|
Make sure Docker Desktop is running. In the `opstrace` project you cloned, run the following command:
|
|
|
|
```shell
|
|
make kind
|
|
```
|
|
|
|
Wait a few minutes while kind creates your Kubernetes cluster. When it's finished, you should see the following message:
|
|
|
|
```plaintext
|
|
Traffic Manager installed successfully
|
|
```
|
|
|
|
Now deploy the scheduler by running the following command in the `opstrace` project:
|
|
|
|
```shell
|
|
make deploy
|
|
```
|
|
|
|
This takes around 1 minute.
|
|
|
|
## Step 2: Create a GitLab application for authentication
|
|
|
|
You must create a GitLab application to use for authentication.
|
|
|
|
In the GitLab instance you'd like to connect with GOP, [create an OAuth application](../integration/oauth_provider.md).
|
|
This application can be a user-owned, group-owned or instance-wide application.
|
|
In production, you would create a trusted instance-wide application so that users are explicitly authorized without the consent screen.
|
|
The following example shows how to configure the application.
|
|
|
|
1. Select the API scope and enter `http://localhost/v1/auth/callback` as the redirect URI.
|
|
|
|
1. Run the following command to create the secret that holds the authentication data:
|
|
|
|
```shell
|
|
kubectl create secret generic \
|
|
--from-literal=gitlab_oauth_client_id=<gitlab_application_client_id> \
|
|
--from-literal=gitlab_oauth_client_secret=<gitlab_application_client_secret> \
|
|
--from-literal=internal_endpoint_token=<error_tracking_internal_endpoint_token> \
|
|
dev-secret
|
|
```
|
|
|
|
1. Replace `<gitlab_application_client_id>` and `<gitlab_application_client_secret>` with the values from the GitLab application you just created.
|
|
Replace `<error_tracking_internal_endpoint_token>` with any string if you do not plan to use error tracking.
|
|
|
|
You can also view [this MR on how to get the token to test error tracking](https://gitlab.com/gitlab-org/gitlab/-/merge_requests/91928).
|
|
You must specify all the parameters when creating the secret.
|
|
|
|
## Step 3: Create the cluster definition
|
|
|
|
1. In your `opstrace` project, run the following command to create a `Cluster.yaml` manifest file:
|
|
|
|
```shell
|
|
cat <<EOF > Cluster.yaml
|
|
apiVersion: opstrace.com/v1alpha1
|
|
kind: Cluster
|
|
metadata:
|
|
name: dev-cluster
|
|
spec:
|
|
target: kind
|
|
goui:
|
|
image: "registry.gitlab.com/gitlab-org/opstrace/opstrace-ui/ gitlab-observability-ui:c9fb6e70"
|
|
dns:
|
|
acmeEmail: ""
|
|
dns01Challenge: {}
|
|
externalDNSProvider: {}
|
|
gitlab:
|
|
groupAllowedAccess: '*'
|
|
groupAllowedSystemAccess: "6543"
|
|
instanceUrl: https://gitlab.com
|
|
authSecret:
|
|
name: dev-secret
|
|
EOF
|
|
```
|
|
|
|
1. Apply the file you just created with the following command:
|
|
|
|
```shell
|
|
kubectl apply -f Cluster.yaml
|
|
```
|
|
|
|
1. Run the following command to wait for the cluster to be ready:
|
|
|
|
```shell
|
|
kubectl wait --for=condition=ready cluster/dev-cluster --timeout=600s
|
|
```
|
|
|
|
After the previous command exits, the cluster is ready.
|
|
|
|
## Step 4: Enable Observability on a GitLab namespace you own
|
|
|
|
Go to a namespace you own in the connected GitLab instance and copy the Group ID below the group name.
|
|
|
|
GOP can only be enabled for groups you own.
|
|
To list all the groups that your user owns, go to the menu in upper left corner and select `Groups`->`View all Groups`. You then see the **Your groups** tab.
|
|
|
|
In your browser, go to `http://localhost/-/{GroupID}`. For example, `http://localhost/-/14485840`.
|
|
|
|
Follow the on-screen instructions to enable observability for the namespace.
|
|
This can take a couple of minutes if it's the first time observability has been enabled for the root level namespace (GitLab.org in the previous example.)
|
|
|
|
Once your namespace has been enabled and is ready, the page automatically redirects you to the GitLab Observability UI.
|
|
|
|
## Step 5: Send traces to GOP
|
|
|
|
[Follow this guide to send traces to your namespace and monitor them in the UI](https://gitlab.com/gitlab-org/opstrace/opstrace/-/blob/main/docs/guides/user/sending-traces-locally.md).
|
|
|
|
## Step 6: Clean up your local GOP
|
|
|
|
To tear down your locally running GOP instance, run the following command:
|
|
|
|
```shell
|
|
make destroy
|
|
```
|
|
|
|
## Known issues
|
|
|
|
### Incorrect architecture for `kind/node` image
|
|
|
|
If your machine has an Apple silicon (M1/M2) chip, you might encounter an architecture problem with the `kind/node` image when running the `make kind` command. For more details, see [issue 1802](https://gitlab.com/gitlab-org/opstrace/opstrace/-/issues/1802).
|
|
|
|
To fix this problem, you first need to create a Dockerfile. Then build and deploy the image:
|
|
|
|
1. Create a new Dockerfile (without a file extension) and paste the following commands:
|
|
|
|
```Dockerfile
|
|
FROM --platform=arm64 kindest/node:v1.23.13
|
|
RUN arch
|
|
```
|
|
|
|
1. Save your Dockerfile, then build the image with the following command:
|
|
|
|
```shell
|
|
docker build -t tempkind .
|
|
```
|
|
|
|
Do not forget the period at the end.
|
|
|
|
1. Create a cluster using your new image with the following command:
|
|
|
|
```shell
|
|
kind create cluster --image tempkind
|
|
```
|
|
|
|
### scheduler-controller-manager pod cannot start due to ImagePullBackOff
|
|
|
|
If while executing `make deploy` in step 1, the `scheduler-controller-manager` pod cannot start due to `ImagePullBackOff`, you must set the `CI_COMMIT_TAG` to a non-dirty state. By setting the commit tag to the latest commit, you ensure the Docker image can be pulled from the container registry.
|
|
|
|
Run the following command to set the commit tag:
|
|
|
|
```shell
|
|
make kind
|
|
export CI_COMMIT_TAG=0.2.0-e1206acf
|
|
make deploy
|
|
```
|