feat(infra): a UAT bring-up that pulls its images instead of building them on the host - #638
Merged
Conversation
… them on the host
`docker-compose.fullstack.yml` cannot deploy. Two properties make it correct on a
developer machine and unusable on the VPS: it builds the Tracker from a sibling
checkout (`../../../evolith_tracker/src`), a path that does not exist where
Coolify clones one repository; and it builds at all, on a 2 vCPU host that would
then be compiling and serving on the same two cores — degrading the environment
for the very people it exists for, precisely while it is updated.
This file builds nothing. Every service names an image CI already published to
GHCR (`ci-cd.yml` here, `images.yml` in the Tracker, both on merge to main), so
what a client exercises is the artifact CI verified rather than a rebuild of the
same source, and a rollback is a pull rather than a rebuild.
Three settings differ from the fullstack file because a host a client can reach
is not a laptop:
- `ASPNETCORE_ENVIRONMENT` is `Staging`, not `Development`, which is what enables
the Tracker's DevBypass auth scheme.
- `CoreApi__MockFallback` is off. Left on, an unreachable core-api is answered
with fabricated evaluations — in UAT, a client believing they saw the product
work.
- `TRACKER_API_UPSTREAM` overrides the image default, which names a Kubernetes
service that does not resolve on this network.
Credentials use `${VAR:?}` rather than defaults: a missing value fails the
bring-up instead of quietly shipping a development password.
Memory limits are declared per service and are not decoration. Node sizes its
heap from the HOST's RAM unless told otherwise, so four Node services on a 7.8 GB
box each plan for memory they cannot have. `mem_limit` caps the container and
`--max-old-space-size` tells the runtime about the cap; without the second the
process meets the OOM killer instead of a GC.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
📊 Bilingual Coverage ImpactPR Changes
Repository Coverage
✅ Good: All EN changes have ES counterparts. Generated by GitHub Actions |
…oting
Measured on the target VPS, not reasoned about. With `start_period: 40s`,
`core-api` did not answer in time, compose aborted every service that depends on
it — agent-runtime, tracker-api and through it the whole Tracker — and the
bring-up ended with four of nine services up and
dependency failed to start: container evolith-uat-core-api-1 is unhealthy
`core-api` then reported healthy on its own and was still healthy fourteen
minutes later. The boot was never the defect; the deadline was. On 2 vCPU this
service starts NestJS and loads the compiled policy bundle before it serves
/health, and the Tracker API runs behind EF migrations on the same two cores.
Both now allow 180s with a 15s interval and 5 retries. Re-run with the images
already pulled: nine of nine up, every healthcheck green, and the full chain
tracker-web -> tracker-api -> core-api answering 200.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds
product/infra/docker-compose.uat.yml— the nine-service stack for the UAT environment on the VPS, pulling published images rather than building them.Why
docker-compose.fullstack.ymlcannot deploy to the VPS, for two independent reasons:../../../evolith_tracker/src). Coolify clones one repository; that path does not exist there.Both properties are correct for a developer machine. Neither survives the move to a host a client reaches.
What changes for the deployed environment
ASPNETCORE_ENVIRONMENTDevelopment(DevBypass auth)StagingCoreApi__MockFallbackfalsefalse, and stated as load-bearing${VAR:?}— missing value fails the bring-upmem_limit+--max-old-space-sizeper serviceThe last row is not tidiness. Node sizes its heap from the host's RAM unless told otherwise, so four Node services on a 7.8 GB box each plan for memory they cannot have;
mem_limitalone caps the container without telling the runtime, which turns a GC into an OOM kill.Verified
docker compose configvalidates.required variable EVOLITH_API_KEY is missing a value, rather than starting.ghcr.io/beyondnetcode/evolith-tracker-web:latestpulls anonymously; the three Core images are private and need a registry credential on the host (out of scope here, it is server configuration).Not in this PR
Creating the Coolify resource, the DNS record for the Tracker web surface, and the registry login on the host — all server-side steps that no repository change can carry.
🤖 Generated with Claude Code