feat: add scaletest load generation infrastructure #15816
Conversation
c10624d to
766ffa0
Compare
| required_providers { | ||
| coder = { | ||
| source = "coder/coder" | ||
| version = "~> 0.23.0" |
There was a problem hiding this comment.
Can we update to a newer version of the provider?
| name = "cli" | ||
| image = "${var.coder_image_repo}:${var.coder_image_tag}" | ||
| command = [ | ||
| "/opt/coder", | ||
| "--verbose", | ||
| "--url=${local.deployments.primary.url}", | ||
| "--token=${trimspace(data.local_file.api_key.content)}", | ||
| "templates", | ||
| "push", | ||
| "--directory=/home/coder/template", | ||
| "--provisioner-tag=scope=organization", | ||
| "--provisioner-tag=deployment=primary", | ||
| "--yes", | ||
| "kubernetes-primary" | ||
| ] | ||
| volume_mount { | ||
| name = "coder-template" | ||
| mount_path = "/home/coder/template/main.tf" | ||
| sub_path = "main.tf" |
There was a problem hiding this comment.
future enhancement idea: could we use the coderd Terrraform provider for this?
There was a problem hiding this comment.
coderd provider can't be used here because it needs a valid coder token up front, and we don't have one since we create coder in the same apply. We've discussed plans to make this possible in the future but it's currently not possible with the provider.
| kubelet_config { | ||
| cpu_manager_policy = "" | ||
| cpu_cfs_quota = false | ||
| pod_pids_limit = 0 | ||
| } |
There was a problem hiding this comment.
Interesting, what made you need to set this?
There was a problem hiding this comment.
I was getting drift on created clusters with this set
| create = local.traffic_types[each.key].job_timeout | ||
| } | ||
|
|
||
| depends_on = [time_sleep.wait_baseline, time_sleep.wait_traffic[each.key]] |
There was a problem hiding this comment.
note for reviewers: these introduce a transitive dependency on the create-workspaces jobs.
| - name: "CODER_PROVISIONER_DAEMON_NAME" | ||
| valueFrom: | ||
| fieldRef: | ||
| fieldPath: metadata.name |
There was a problem hiding this comment.
Is this something we should add by default to the coder-provisioner Helm chart?
There was a problem hiding this comment.
I think it could be useful as a default as long as you can override it
Closes coder/internal#149
This creates workspaces in each region and runs traffic generation against the workspaces colocated in the region.