loki fix log retention and compactor

[DaMandal0rian]

PR Type

Enhancement

---

Description

• Enable compactor retention and delete settings

• Set overall log retention to 30 days

• Upgrade Grafana Loki, Promtail, and Mimir images

---

Changes walkthrough 📝

	Relevant files
Configuration changes	loki-config.yml Configure Loki compactor retention settings                            logging/conf/loki-config.yml Enabled compactor log retention Added delete delay and worker count Set limits_config retention period Clarified retention comment in table_manager +6/-1
Dependencies	docker-compose.yml Upgrade Loki ecosystem images                                                        logging/grafana-loki/docker-compose.yml Bumped Loki image to 2.9.14 Updated Promtail image to 2.9.14 Updated Mimir image to 2.9.14 +3/-3

---

Need help?

Type /help how to ... in the comments thread for any questions about PR-Agent usage.

Check out the documentation for more information.

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🧪 No relevant tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Retention Consistency Verify that limits_config.retention_period aligns with reject_old_samples_max_age to avoid unintended data rejection before retention. reject_old_samples_max_age: 168h retention_period: 720h # Retain logs for 30 days Compactor Delete Workers Assess if setting retention_delete_worker_count to 150 is appropriate for resource usage and ensure it won’t overload the system. retention_delete_worker_count: 150 delete_request_cancel_period: 5m Image Version Bump Confirm that upgrading Loki, Promtail, and Mimir to 2.9.14 is compatible with existing configurations and cluster components. image: grafana/loki:2.9.14 ports: - "3100:3100" command: -config.file=/etc/loki/loki-config.yaml labels: - traefik.enable=true - traefik.http.services.loki.loadbalancer.server.port=3100 - traefik.http.routers.loki.rule=Host(`logging.subspace.network`) - traefik.http.routers.loki.tls.certresolver=le - traefik.http.routers.loki.entrypoints=websecure - traefik.docker.network=traefik-proxy - traefik.http.routers.loki.middlewares=loki-auth - traefik.http.middlewares.loki-auth.basicauth.usersfile=/etc/traefik/loki-auth.txt volumes: - /root/loki/conf/loki/etc/loki-config.yaml:/etc/loki/loki-config.yaml:ro networks: - loki - traefik-proxy promtail: image: grafana/promtail:2.9.14 command: -config.file=/etc/promtail/config.yml ports: - 127.0.0.1:3200:3100 networks: - loki mimir: image: grafana/mimir:2.9.14 volumes:

[github-actions]

PR Code Suggestions ✨

Explore these optional code suggestions:

Category	Suggestion	Impact
Possible issue	Move retention to table_manager The limits_config block does not support retention_period, so this setting will be ignored. Remove it from limits_config and place retention_period: 720h under the table_manager section where retention is actually enforced. logging/conf/loki-config.yml [50-53] limits_config: reject_old_samples: true reject_old_samples_max_age: 168h - retention_period: 720h # Retain logs for 30 days + # retention_period moved to table_manager Suggestion importance[1-10]: 8 __ Why: The retention_period is not supported under limits_config and will be ignored, so moving it to the table_manager block ensures retention is actually enforced.	Medium
General	Lower delete worker count A worker count of 150 can overwhelm resources during retention deletion. Lower this to a more conservative value (e.g., 50) or rely on the default to prevent spikes in CPU/memory usage. logging/conf/loki-config.yml [42-48] compactor: working_directory: /loki/boltdb-shipper-compactor shared_store: filesystem retention_enabled: true retention_delete_delay: 2h - retention_delete_worker_count: 150 + retention_delete_worker_count: 50 # reduced to avoid resource exhaustion delete_request_cancel_period: 5m Suggestion importance[1-10]: 4 __ Why: Reducing retention_delete_worker_count helps prevent CPU/memory spikes but is a minor operational tuning rather than a bug fix.	Low