Operations

Monitoring and notifications

Two things: numbers you go and look at, and messages that come and find you. The metrics are request and resource figures for what you publish; the notifications are the rules that decide what is worth interrupting you for.

Metrics

The overview shows, over a window of 60 seconds, 5 minutes, 1 hour or 24 hours: requests and requests per second, the error rate as 4xx plus 5xx, average response time, average CPU and memory, and traffic in and out. Below that, one row per stack with the same figures and a trend line; a stack's own page repeats them per service, with charts of responses by status class and of CPU and memory.

A stack that has just been deployed or has no inbound traffic says so rather than drawing an empty chart.

PlanMetrics and history kept
Community3 days
Personal7 days
TeamUnlimited
MaxUnlimited

Notification channels

A channel is a place messages go: Discord, Slack, Telegram, Microsoft Teams, a generic webhook, or email. Each takes what that service needs - a webhook URL, or a bot token and chat ID, or an SMTP host, port, sender and recipients - and can be tested from its own card.

Email is available on every plan. The other channels are part of Team and above.

What a channel is told about

Events are chosen per channel, in five groups:

The defaults are chosen so that a new channel is useful without being noisy: failures and things going down are on, successes and threshold alarms are off until you turn them on.

The panel's own outbound mail

Separate from the per-channel email alerts, the panel has one SMTP server it uses for transactional mail - temporary passwords and password resets. It is configured on its own page next to Notifications, and is filled in for you if you turn on the mail server. Without it, password resets go nowhere.

dev