Thoughtworks · Client engagement
Costa
Application monitoring and alert delivery
What I built
An application monitoring system with log alerts, team notifications and homepage checks.
01 / Monitoring flow
Application logs into team notifications.
I worked as a DevOps Engineer on Costa’s digital transformation project, using AWS, Terraform and monitoring tools including Prometheus, Graylog and Alertmanager.
I developed an application monitoring and alerting system for the team. I connected Graylog log alerts with Alertmanager and Slack, and added scheduled checks of the website’s homepage.
The system gave the team a central place to review current alerts and a way to receive notifications in its everyday collaboration channels. Log context and environment labels helped connect each notification to the application problem behind it.
Monitoring paths
Log-based alerts
- Application logsLog events
- GraylogStreams & conditions
- Notification integrationAPI callback
- AlertmanagerAlert handling
- SlackTeam notifications
Scheduled website checks
- Homepage checksScheduled AutoTest
- SlackCheck reports
A concrete alert rule
I configured a Graylog stream to select ERROR-level messages. A
message-count condition triggered an alert when at least one matching message appeared
within a five-minute window. Graylog’s alert configuration included the last message,
giving the alert additional context.
A notification callback sent the alert to Alertmanager over HTTP. The integration carried an alert name, environment labels and a Graylog link in its annotations. Alertmanager’s configuration and templates supported delivery to Slack.
02 / Team response
Context and controls for investigation.
I set up separate Slack notification channels for the China and Japan business environments. Alert messages included the environment, alert name and a Graylog link, with distinct firing and resolved states.
- Inspect
- View the current alert, its trigger time, stream and rule details, then follow the log link to investigate.
- Silence
- Pause matching notifications by alert name and environment for a defined period while handling an issue.
- Resume
- Review active silence rules and end a silence early when the team is ready to receive notifications again.
These controls gave the team a practical workflow for receiving an alert, finding the relevant context and managing notifications during investigation.
03 / Homepage checks
A view of the website’s availability.
The AutoTest notifier checked the homepage URL on a schedule and sent reports to Slack. This added a view of the website’s entry point alongside the application’s log-based alerts.
I delivered a system that combined error detection, alert context and team notification, with controls for investigating issues and scheduled checks for observing the website. The team could see both kinds of signal within Slack.