Azure Batch Java Job Orchestration
Run and manage large-scale parallel and HPC workloads on Azure from Java.
The skill recommends Microsoft Entra ID and keeps shared-key material in environment variables; no credential hardcoding, malware, or covert exfiltration is shown. However, it includes pool/job/node deletion, administrator tasks, remote-login settings, and SAS URLs without confirmation gates, least-privilege guidance, data-flow disclosure, cost warnings, or rollback instructions, so points are deducted.
The document is organized and covers synchronous, asynchronous, long-running, and error-handling patterns. Reliability is reduced because it depends on beta.5, file-contained tests do not reproduce the API signatures/imports/model usage, references/examples.md visibly ends mid-example, and abnormal-input or diagnostic guidance is limited; static calibration keeps this at or below 10.
The name, audience, Azure Batch scenario, trigger phrases, and prerequisites are fairly clear, covering pools, jobs, tasks, nodes, and schedules. It lacks explicit non-fit boundaries, input/output contracts, compatibility limits, and Chinese-language support; Azure reachability, regional availability, and mainland-China network constraints are also not addressed.
The skill has front matter, MIT licensing, author and version metadata, layered installation/authentication/concepts/operations sections, and reference links. It lacks a changelog, explicit maintenance owner/update path, compatibility matrix, FAQ, and known-limitations section, with some duplication between the main document and references.
The examples cover common Azure Batch lifecycle operations and Java integration patterns and could support a basic implementation. Static evidence does not verify execution; beta dependencies, platform-specific commands, cloud cost exposure, and permission risks mean substantial review may still be needed, so the score remains within the static maximum of 7.
The skill supplies Maven, GitHub, API, product-documentation, and sample links, while repository context shows general CI and tests. There is no skill-specific acceptance suite, committed execution result, or independent corroboration, and no code was executed in this review, so evidence is limited.
- Treat pool/job deletion, task termination, administrator tasks, autoscaling, and remote-login examples as high-impact operations requiring explicit user confirmation.
- Compile-check the beta.5 dependency and example API usage against the target Java SDK version, and add failure diagnostics, permission scope, cost, and rollback guidance.
- Clarify sensitive-data handling, storage access scope, and credential lifetimes for SAS URLs, task inputs/outputs, and stdout/stderr.
- Mainland-China users should separately verify reachability of the selected Azure Batch region, identity endpoints, Maven repository, and documentation links.
What it does & when to use it
This skill helps Java developers use the Azure Batch SDK to operate large-scale parallel and high-performance computing workloads. It covers pools, jobs, tasks, compute nodes, and recurring job schedules, with synchronous and asynchronous client examples. It also documents Microsoft Entra ID and shared-key authentication, environment variables, long-running operations, and Batch-specific error handling. It fits teams that already have Azure Batch access and need concrete Java patterns for submitting and monitoring compute work.
Shows how to add the azure-compute-batch Maven dependency, configure the Batch endpoint and authentication variables, and create synchronous or asynchronous Batch clients; create, inspect, list, resize, autoscale, and delete pools; create, inspect, batch-submit, terminate, and delete jobs and tasks; retrieve task output; list or reboot nodes and retrieve remote-login settings; create and inspect job schedules; and inspect BatchErrorException codes and messages.
- A Java team needs to run many independent workloads in an Azure Batch pool and wants collection-based task submission.
- An HPC workload needs pool capacity to follow pending-task volume and uses an autoscale formula to adjust nodes.
- A service needs operational progress reporting and reads active, running, and completed task counts for a job.
- A production deployment wants Entra ID or managed identity authentication instead of managing a shared key.
- A batch pipeline needs maximum wall-clock time, retry limits, or exit-code-driven job actions for task failures.
Pros & cons
- Covers the core Batch resource model: pools, jobs, tasks, nodes, and job schedules.
- Provides examples for Entra ID, managed identity, shared keys, synchronous clients, and asynchronous clients.
- Includes practical patterns for bulk task submission, autoscaling, long-running operations, task counts, and error handling.
- Provides Maven coordinates, environment variables, and concrete Java types for starting implementation.
- The documented package version is 1.0.0-beta.5, and the source does not state its stability or upgrade compatibility.
- It assumes an existing Azure Batch account, subscription, and compute pool rather than provisioning the full environment.
- It focuses on Batch data-plane operations; it recommends the management SDK for pool management but does not provide its implementation.
- The source provides no standalone test results, cost estimate, or performance benchmark for this individual skill.
How to install
Add this dependency to the project's Maven configuration: <dependency><groupId>com.azure</groupId><artifactId>azure-compute-batch</artifactId><version>1.0.0-beta.5</version></dependency>. For installing the Agent Skill collection, the README provides: npx skills add microsoft/skills. Select the required skill in the wizard and install it into the target agent directory. The source does not provide a separate installation command for this individual skill.
How to use
After installation, use a direct coding request such as “Use the Azure Batch SDK for Java to create a pool and submit parallel tasks” or “Help me configure BatchClient Java authentication.” Have an Azure Batch account, an Azure subscription, and a pool configured with compute nodes. Set AZURE_BATCH_ENDPOINT; shared-key authentication also requires AZURE_BATCH_ACCOUNT and AZURE_BATCH_ACCESS_KEY.
Compared to similar skills
For authentication, the skill recommends Microsoft Entra ID and presents shared-key credentials as an alternative. For pool management, it points to the azure-resourcemanager-batch management SDK as an option supporting managed identities.