Cloud infrastructure provider DigitalOcean recently released capabilities for monitoring servers and sending alerts. While not novel, this free feature is indicative of growing industry attention paid to server and application insight.
The DigitalOcean Monitoring service is agent-based. It works with both new and existing virtual machines (called "droplets"). Developers install the agent on Linux-based droplets during provisioning, or by running a post-build script. The agent captures a variety of metrics, including: CPU, memory utilization, disk I/O, disk utilization, inbound bandwidth, and outbound bandwidth. These metrics are only visible via dashboards, and not accessible via API. Because metrics are captured via an on-box agent, information like "top processes" is also available. Developers create alert policies to raise alarms when performance thresholds get breached. These policies apply to a single droplet, or group of droplets. Alert notifications can go either an email address or Slack channel. DigitalOcean stores metrics data for up to 30 days. But, if a user destroys a droplet, all metric data is immediately destroyed as well.
Before launching their Monitoring service, DigitalOcean relied on partner offerings to fill the gap. DigitalOcean co-founder Mitch Wainer told journalist Ben Kepes that this functionality didn't come as a surprise to their key partners.
Of course, the obvious question for Wainer is what this means to existing DigitalOcean partners. Wainer was pretty open and honest, admitting that at least at the base level, this might impact partners. But he quickly added that DigitalOcean has been very open about its intentions, giving these vendors many months to prepare for the change.
Other major IaaS vendors also support a wide range of monitoring partners, but have offered built-in monitoring for years. Amazon CloudWatch has a free tier for server monitoring and can store metric data for over a year. Additionally, AWS offers visualization, alerting, API access, a more advanced tier (for a cost), custom metrics, and log monitoring. Microsoft also offers comprehensive monitoring for Azure virtual machines. Users can set up alerts, archive metrics data to long-term storage, consume and query via API, and more. Google Stackdriver monitoring keeps an eye on both Google Compute Engine and AWS servers. It has a free tier as well as a more advanced tier of service. Like the other IaaS providers, Google offers custom metrics, alerting policies, rich notification options, API access, and more.
User expectations for server and application monitoring are changing. A report on The New Stack points out how containerization impacts monitoring needs. Ephemeral services, proliferation of "things" to track, a focus on "services" not "servers", and new types of end users have all rendered the classic monitoring approach inadequate. Ephemeral services means that vendors need a new approach to pricing that takes into account service instances that come and go. To handle troubleshooting more complex application topologies, AWS introduced X-Ray in 2016. There's been an explosion of tools and services to help modern teams monitor services and troubleshoot problems. Existing vendors like New Relic, Dynatrace, and Datadog are evolving to meet the need. Emerging solutions like Prometheus, Pivotal Cloud Foundry Metrics, Honeycomb, and Zipkin have industry attention.
DigitalOcean Monitoring is free to use, and available in any of their eight locations around the world. At the end of the blog post announcing the service, DigitalOcean outlined future plans for the service.
With the first iteration of our Monitoring service out the door, we're already working on what's next. Some features you will see soon include: