Creează un cont pentru a te abona la acest blog și a-i urmări articolele pe pagina Lectură.
-
Give a coding agent write access to your repo and the default GITHUB_TOKEN quietly hands it whatever else that token can reach.
-
Three teams share one Azure OpenAI deployment and finance wants a per-team number.
-
Flux can reconcile a vLLM ServingRuntime and an InferenceService just fine, but the model weights it points at aren't tracked by Git.
-
Five AKS platform decisions that quietly determine what your GPU capacity costs — workload placement, driver ownership, partitioning, scale-to-zero, and measuring utilization instead of allocation.
-
Write-only fields like service connection passwords produce a diff on every plan.
-
Why terraform plan reports changes you never made — seven causes, from API normalization to jsonencode formatting — and six ways to handle them, with a framework for choosing between suppressing and fixing.
-
What WebAssembly and WASI actually offer outside the browser — cold start times, footprint, real multi-arch builds and capability-based security — and where Wasmtime, Wasmer, WasmEdge and Spin fit.
-
When a shared kernel stops being an acceptable trade: what a microVM is, how Firecracker, gVisor and Kata differ, the workloads that justify them, and the cases where a plain container is still right.
-
What Terraform state buys you and what it costs — dependency tracking and cross-team outputs on one side; lock conflicts, drift, and plans that slow to a crawl at scale on the other.
-
Running a shared Azure DevOps template repo: versioning with git tags, communicating breaking changes, tracking which pipelines consume what, and choosing between task, job and stage level templates.
-
Setting up a themed PowerShell prompt with oh-my-posh — the Nerd Font, the winget install, Terminal-Icons, and wiring a theme into your profile so it survives a new window.
-
Pointing the Windows podman client at a rootless Podman already running in your existing WSL instance over SSH, instead of letting Podman Desktop stand up a second WSL distro beside your dev environment.
-
How Linux networking got from ifconfig and a one-line resolv.conf to iproute2, NetworkManager, systemd-networkd and systemd-resolved — what each layer solved, and why the simple version stopped being enough.
-
Moving files between repositories without losing their history, using git filter-repo — why it replaced filter-branch, and the four steps from cleaning the source repo to merging with --allow-unrelated-histories.
-
A first look at Azure Bicep as a replacement for ARM templates — installing the CLI through az, decompiling an existing template to compare the two side by side, and deploying it straight to a resource group.
-
Building a Kubernetes cluster on Raspberry Pis with k3s — why k3s suits resource-constrained hardware, preparing the SD cards and boot configuration, then installing the server and joining agent nodes.
-
Upgrading an Ubuntu WSL instance to 20.04 LTS with do-release-upgrade, including the -d flag you need when it insists there is no new release available.
-
Managing kubectl contexts once they outgrow a single config file — switching temporarily with KUBECONFIG, and merging separate context files into one with kubectl config view --merge --flatten.
-
Bringing a fork back in line with the repository it came from — adding the upstream remote, fetching it, and merging into your local master before pushing.
-
The two commands worth remembering for submodules: pulling them down for the first time, and updating them with upstream changes.
-
Restoring a large backup on Windows ends in 'Error 2006: MySQL server has gone away'.