What is Max. Failed Connection Retries used for?

Short answer

Max. Failed Connection Retries in the vService Limits profile determines how many times the connection is retried to the same server when a connection to the Backend Service cannot be established; the default is 3, the range is 0–1,000. Each attempt is limited by Connect Timeout. If the attempts are exhausted, the request fails and the client receives 503. The field does not disable the server and does not move the request to another server; removing the server from traffic is the job of the Health Checks profile.

How Does It Work?

1. TR7 tries to connect to the selected Backend Service server; each attempt is limited by Connect Timeout (default 20 s).
2. If the connection cannot be established, it is retried to the same server as many times as the set number.
3. When the attempts are exhausted, the request ends with 503. On a server that does not respond, in the worst case the client can wait the Connect timeout duration back to back as many times as the total of the first attempt and the retries (with the defaults, 4 × 20 = 80 seconds).

Max. Failed Connection Retries in TR7

• Setting: Load Balancer → Profiles → vService Limits → right-click on the profile row → Edit (or Add) → Max. Failed Connection Retries → Save; the profile is selected in the Details → vService Limits box of the vService. The value is applied without interruption; the vService does not restart. In a cluster, it is also sent to the peer device with the default auto-sync; if Manual Sync Mode is on, it is sent with Synchronize.
• vService type: HTTP and L7 TCP.
• Monitoring: The retries are seen in Live Stats on the per-second Retries card in the Backend Services group and in the Backend Services table; for individual requests, while the HTTP vService is in Debug mode, in the Retries column in Debug Mode Logs (0 = connected on the first attempt).
• Recommendation: In API services, keeping it low is recommended so that the error is seen quickly (API services).
• WEB-CLI: lb limits update id <id> --maxRetries 5 (limits examples).

Common Misconceptions

• "A request that returns a 5xx response is retried." The field works only on connection setup errors; the error responses returned from the server are not retried.
• "The attempts are made to another server." The retries are made to the same server.
• "After a few failed attempts, the server goes down." The status of the server is determined by the Health Checks profile (What does a Health Check do; what happens if I don't assign one?).

How it is done in the documentation

• vService Limits · Max. Failed Connection Retries
• Timeouts · Connect Timeout
• Live Stats · Frontend

Related questions

• What is the vService Limits profile; what are the default limits?
• What does a Health Check do; what happens if I don't assign one?
• What do the values in the Timeouts profile (HTTP Keepalive Timeout, HTTP Request Timeout, Connect, Server, Client, Queue, Tunnel(Websocket) Timeout, Client FIN, Server FIN) mean?