Mackerel vs Prometheus

Mackerel is a cloud-based SaaS service for monitoring servers and team-managed infrastructure. Installing its agent starts monitoring, after which hosts can be grouped into services and roles. It visualizes host metrics and user-defined service measures, including response time, sales and page views. Monitoring features include combining metrics, tracking predicted values and machine-learning anomaly detection. Cloud integrations cover Amazon Web Services, Microsoft Azure and Google Cloud; a container agent treats Amazon ECS tasks and Kubernetes pods as hosts. Users can extend monitoring with open-source plugins, APIs and the mkr command-line tool. Customer Reliability Engineers can offer technical guidance on implementation and use, while standard-plan technical support is email-only. The free plan has five standard and micro hosts total, 500 metrics, one-day retention, 10 monitoring rules, one dashboard and up to 5M trace spans. A two-week trial includes Standard-plan functions and a 100M-span allowance. Standard host usage costs 2,180 yen, tax included, per host month.

Prometheus is a free, open-source toolkit for systems monitoring and alerting that collects metrics as time series. Each series is identified by a metric name and key-value labels, and PromQL is used to query the data. Prometheus typically gathers metrics by pulling over HTTP; it can also receive pushed data through an intermediary gateway. Targets can be found with service discovery or configured statically. The main server runs as a standalone monolithic binary without external dependencies, and official components are available as precompiled binaries and Docker images. Its ecosystem includes client libraries, exporters, Pushgateway and Alertmanager, with Grafana for visualization and webhooks for alert notifications. Prometheus and most exporters support TLS, client certificate authentication and HTTP Basic Authentication. The project advises against exposing component HTTP endpoints to public networks without protective measures. Servers do not depend on network storage or remote services, which the project says can help during outages. Prometheus is not recommended for uses requiring 100% accurate data, such as per-request billing.