What are the recommended values for slow-responding services (databases, reporting)?

Short answer

Timeout is selected comfortably above the service's normal response time under load but below 120 seconds (e.g. 10 s); Period 5–10 seconds, Required Failures 3 and Required Successes 2–3 are a good starting point; the docs also recommend increasing the Timeout and Required Failures values for slow services (Services that respond slowly). A lightweight health address or query is used instead of heavy report queries. With these settings (Period 5, Timeout 10, Required Failures 3), a server that does not respond is removed from traffic in about 40–48 seconds on HTTP and L7 TCP vServices.

How Does It Work?

Field	Recommendation	Reason
Timeout	1.5–2 times the worst normal response time observed (e.g. 10 s), below 120 s	Slow but healthy responses are not counted as failed
Period	5–10 s	No unnecessary load is put on the server
Required Failures	3	A single slow response does not take the server down
Required Successes	2–3	The server does not return to traffic before it becomes stable

Example calculation (Period 5, Timeout 10, Required Failures 3, Required Successes 2; formula: How many seconds does it take for a server to be considered "down", and how do I calculate it?):

• Removal: On a server that does not respond, each check takes 10 seconds, followed by a 5-second wait; three failed checks take about 40–45 seconds, and the effect on traffic adds at most 2.5 seconds.
• Return: When the server responds again, two successful checks take about 5–10 seconds, and the effect on traffic adds at most 2.5 seconds.

In TR7

• Values: Load Balancer → Profiles → Health Checks → profile → Edit (How should I set the Period, Timeout, Required Successes and Required Failures values?). If Period and Timeout are set to 120 seconds or more, 1 second is applied.
• Oracle database: A lightweight command (e.g. a query that returns a single row) is run with Oracle Scenario in the Oracle type.
• Other databases: TCP, or TCP Scenario, which sends data suited to the protocol and waits for a response.
• Reporting and web-based services: An HTTP check to a /health address that does not generate a report and only returns that the service is running.
• Monitoring: The measured duration of the check is seen as a chart in the Health Check tab on the Dashboard; Timeout is selected above this duration.

Common Misconceptions

• "If I keep the Timeout very large, I will be safe." A value of 120 seconds or more is applied as 1 second; in addition, noticing a failure takes longer.
• "The health check should run the real report query." Heavy queries strain the server and slow down the check; a lightweight check is sufficient.
• "Keeping the Period low is safer." On slow services, frequent checks add load; stability is determined by the thresholds.

How it is done in the documentation

• Health Checks · Services That Respond Slowly (Database, Reporting)
• Health Checks · Recommended Values and Tuning Guide
• Health Checks · Timeout Behavior
• Health Checks · Oracle Scenario

Related questions

• How should I set the Period, Timeout, Required Successes and Required Failures values?
• How many seconds does it take for a server to be considered "down", and how do I calculate it?
• Which algorithm is recommended for database connections?