PricingPricingDocsDocsBlogBlogBook a demoBook a demoCompanyCompany (opens in a new tab)
Product
DocsAPI ReferencePricingConsole DemoBlogCompany
Start Freev3.4.7
Platform & Observability

Alerts

Alert rules you write as JavaScript expressions, checked every minute, sent to Slack or an HTTP webhook, and logged for 30 days.

The challenge

CPU stays above 80% for 5 minutes, and no one gets a message

Wave Alerts docs (the default rule for workload metrics); prometheus.io, Alerting overview; Prometheus and Alertmanager configuration docs.

Kubernetes has no built-in way to send a message when a workload stays hot. The usual setup is Prometheus with Alertmanager: "Alerting rules in Prometheus servers send alerts to an Alertmanager" (prometheus.io).

That is two more systems to run and configure. By default, Prometheus scrapes and evaluates rules every 1m, and Alertmanager waits 30s (group_wait) before the first notification of a group.

The rule in the title is the Wave default for workload metrics: send a message when, over at least 5 minutes, the maximum CPU or the average memory reaches 80%.

A quiet channel, then a message when the rule matchesWithout WaveWith Waveno messageSlack message sent
Without Wave
CPU stays above 80%, and the channel stays quiet80%5 min above 80%0510 min#ops-alertsNo new messages
CPU stays above 80% for more than 5 minutes. Kubernetes records the metrics, but nothing sends a message.
With Wave
The default rule matches, and Slack gets a message80%5 min above 80%0510 min#ops-alertsHigh CPUWorkload: prod/checkoutCPU Usage: 88%
The default workload rule, at least 5 minutes with CPU at or above 80%, matches on the next 1-minute check. Wave fills in the template and sends it to the Slack channel.

Default rule, 1-minute check, and message template: Wave Alerts docs. The CPU series, the workload name, and the channel are illustrative; 88% is the max of the window, as in the docs example.

How Wave does it

From a CPU spike to a message in Slack
Prometheus + AlertmanagerTwo more systems to run
  1. Prometheus
    scrapes metrics every 1m by default
  2. ↓rules→
    Alerting rule
    a PromQL expression, evaluated every 1m
  3. ↓fires→
    Alertmanager
    groups alerts, waits 30s first
  4. ↓routes→
    Receiver
    chat, webhook, and more
Wave AlertsBuilt in, logged for 30 days
  1. Events
    workload metrics, schedule phases, Autopilot logs
  2. ↓match→
    Rule
    a JavaScript expression, checked every minute
  3. ↓render→
    Message
    your template, with the workload and its CPU
  4. ↓send→
    Channel
    Slack Webhook, Slack Web API, or HTTP
Default lane: prometheus.io (Alerting overview, configuration) and the Alertmanager configuration docs. Wave lane: Wave Alerts docs.
1

Add a channel

Send to a Slack Webhook, the Slack Web API, or an HTTP webhook. One channel can serve many rules, and each channel can go through an HTTP proxy.

2

Write the rule

A rule is a JavaScript expression over one of three event types: workload CPU and memory, Deployment scheduling phase START and END, or missing Autopilot logs. Wave checks rules every minute.

3

Check every delivery

Each send is logged with its response and status, and logs are kept for 30 days. Wave does not retry a failed send, so a failure does not turn into a flood of messages.

Related in Platform & Observability

Programmable API →SSO (OIDC) →PV Auto-Expansion →PV Cleanup →Cluster Map →Prometheus & Grafana Integration →Read the docs →
Book a demoStart free
Product
Wave GPUAI Infrastructure Professional ServiceWaveWave AI Agent2026 Q4
Resources
DocsAPI ReferenceBlogConsole DemoPricing
Company
STCLabBook a demoContact
STCLab© 2026 Wave All rights reserved.
Terms of ServicePrivacy PolicyCookies