Skip to main content

Stuck gocipher

Description​

This alert fires when a single gocipher instance is persistently failing to send its captured TLS secrets to the server, while the rest of the fleet is healthy.

A gocipher sends secrets over a pooled keep-alive HTTP connection to the hub. If a pooled connection goes half-dead (for example a proxy in front of the hub silently closes its side), that one instance can keep reusing the dead connection and receive repeated 502 Bad Gateway responses (GCPH-OUT-204), dropping its secrets until the pod is restarted.

Because the failure is confined to one pod out of potentially dozens, its handful of errors is easily hidden in the fleet-wide tooManyLogs noise. This probe inspects the errors per instance, so a single wedged gocipher is surfaced on its own.

info

The gocipher retries each send on a fresh connection before logging GCPH-OUT-204, so a single sustained occurrence means that instance could not reach the server on any connection — its decrypted traffic is at risk.

Default configuration​

"stuckGocipher": {
  "active": true,
  "minErrorsPerMin": 0.2,
  "delayWhenInactive": "PT5M",
  "delayWhenActive": "PT1M"
}
  • minErrorsPerMin — an instance is considered stuck when it produces at least this many secret-send failures per minute over the period (and at least two failures in the window, to avoid firing on a single transient hiccup).

Mail​

Content​

The mail lists every gocipher instance considered stuck, with its host, the number of failures and the rate over the period.

Sample​

{
  "endpoint": "https://...",
  "name": "stuckGocipher",
  "status": "ACTIVE",
  "since": "2026-06-19T08:30:25.969Z",
  "stuckInstances": [
    {
      "instance": "local-gocipher-wxx65",
      "host": "ip-192-168-98-185.eu-west-1.compute.internal",
      "count": 12,
      "ratePerMin": 2.4,
      "statusCode": 502,
      "msg": "Error received from server while sending secrets"
    }
  ],
  "windowMin": 5
}