Page 8 - Docker

Service discovery is the invisible backbone of every Docker Swarm cluster: internal DNS, virtual IP addresses and a cluster wide routing mesh make sure containers reach each other through stable names, even though the actual placement can change at any time. Understanding these mechanisms lets you resolve networking problems in Swarm much faster.

GitLab Auto DevOps generates a complete build, test and deploy pipeline the moment it is enabled, with no custom .gitlab-ci.yml required. For stateless microservices that works remarkably well, but for a complex system like Magento with Composer dependencies, database migrations and multi-language static content deployment, the automation hits its limits quickly. This article shows how Auto DevOps works internally, exactly where it struggles with Magento, and what a hybrid approach combining generated and custom jobs looks like.

Debugging PHPUnit tests in a Docker environment with PhpStorm is one of the most frustrating parts of PHP development. Configuration usually fails because of XDEBUG_MODE, incorrect path mapping, or a remote interpreter that was never set up correctly. This guide shows the complete path, from installing Xdebug to running coverage reports.

Reproducible Docker builds produce the same image digest from identical source code and identical configuration on every build, no matter when or on which machine the build ran. SOURCE_DATE_EPOCH eliminates the biggest source of noise behind non-determinism, the current timestamp embedded into every layer.

GitLab Pages turns static documentation, whether generated from PHPDoc, MkDocs, or another generator, into a fixed part of the CI pipeline and republishes it automatically on every merge, without anyone having to upload files by hand.

An open database port does not mean MySQL is ready to process connections. depends_on with condition service_healthy ties the startup order of Docker Compose services directly to the healthcheck status, instead of relying on a plain reachability signal, and eliminates one of the most common sources of errors in multi service stacks.

An ephemeral review app is automatically created as soon as a pull request is opened, and disappears without a trace once it is merged or closed. Docker Compose builds a dedicated, isolated stack per branch for this, reachable through a dynamic subdomain, with no manual environment management at all.

Switching between editor and terminal to run Composer or npm costs focus and context. PhpStorm can integrate Composer, Node.js, npm and Yarn directly, with autocompletion for composer.json and package.json, script buttons and a full output view right in the IDE.

When PHPStan is green locally but red in the CI pipeline, it comes down to a configuration mismatch. The problem is solvable: align pre-commit hooks, PhpStorm quality gates and the CI pipeline so that what runs locally checks exactly what the pipeline checks, no push that fails right out of the gate.

Before Docker officially supported named volumes, the community solved the problem of persistent data with a workaround: a dedicated, never running container that only declared data directories and passed them on to other containers via --volumes-from. This data-only container pattern was standard for years, but is now replaced by named volumes in the vast majority of cases, with only a few remaining niches where the old approach still plays a role.