Feature Request: Add configurable concurrency control for parallel:matrix jobs
Everyone can contribute. Help move this issue forward while earning points, leveling up and collecting rewards.
Problem Statement
When using parallel:matrix to create multiple job instances, GitLab creates and attempts to run ALL jobs simultaneously. For example, a matrix generating 100 jobs will queue all 100 at once, potentially overwhelming runners and external resources.
Proposed Solution:
Add configurable concurrency control in the CI YAML to limit how many parallel matrix jobs can run simultaneously:
deploy:
parallel:
matrix:
- ENVIRONMENT: [dev, staging, prod1, prod2, ..., prod50]
max_concurrent: 5 # Only 5 jobs run at a timeUse Cases
- API rate limiting: Prevent jobs from hitting external service limits
- Resource management: Control infrastructure load without modifying runner configuration
- Cost control: Limit concurrent cloud runner usage
- Database connections: Prevent too many simultaneous connections
Benefits
- Job-level concurrency control without affecting other pipelines
- Better resource utilization
- Prevents job failures due to resource exhaustion
- More predictable pipeline behavior
Alternative Considered
Currently users must use workarounds like runner-level limits (affects all jobs), resource groups (only 1 concurrent), or manual stage batching (defeats the purpose of matrix).
Edited by 🤖 GitLab Bot 🤖