Runtime

[runtime]

​

Background work

​

background_interval

Sets how often background services check for pending work.

• Default: 1 s
• Type: duration

Example { secs = 1, nanos = 0 } runs the next background check about one second later. It is not a workload schedule.

​

Retry policy

​

retry.strategy

Chooses how the wait between retries grows.

• Default: Exponential
• Type: choice: Constant, Exponential, Fibonacci

Example Exponential backoff can wait 100 ms, then 200 ms, then 400 ms, until its configured maximum delay. Configure one strategy at a time.

​

retry.strategy.Constant.delay

Sets the fixed wait between retries for the Constant strategy.

• Default: Set when used
• Type: duration

Example { secs = 1, nanos = 0 } waits one second each time rather than increasing the delay.

​

retry.strategy.Exponential.min_delay

Sets the shortest wait used by this retry strategy.

• Default: 100 ms
• Type: duration

Example A 100 ms minimum lets the first retry happen quickly, while later retries can wait longer according to the selected strategy.

​

retry.strategy.Exponential.max_delay

Caps how long the retry strategy waits between attempts.

• Default: 5 s
• Type: duration

Example A five-second maximum stops backoff from growing beyond five seconds between retries.

​

retry.strategy.Exponential.factor

Sets how much an Exponential retry delay grows after each retry.

• Default: 2.0
• Type: number

Example A factor of 2.0 can grow a 100 ms delay to 200 ms, then 400 ms, until max_delay is reached.

​

retry.strategy.Fibonacci.min_delay

Sets the shortest wait used by this retry strategy.

• Default: Set when used
• Type: duration

Example A 100 ms minimum lets the first retry happen quickly, while later retries can wait longer according to the selected strategy.

​

retry.strategy.Fibonacci.max_delay

Caps how long the retry strategy waits between attempts.

• Default: Set when used
• Type: duration

Example A five-second maximum stops backoff from growing beyond five seconds between retries.

​

retry.max_retries

Limits the number of retries allowed by the shared retry policy.

• Default: 3
• Type: integer

Example 3 bounds the retries after an operation asks to be retried. It does not bypass checks for whether a write already succeeded.

​

retry.jitter_seed

Optionally makes the random variation in retry timing repeatable.

• Default: Unset
• Type: integer; optional

Example A fixed seed is useful when reproducing a test. Leave it unset when you do not need the same sequence of retry delays.

​

Retry strategies

runtime.retry.strategy.Constant uses one delay. Exponential grows between min_delay and max_delay using factor. Fibonacci grows between the same bounds using its sequence. Keep only one strategy in a file. Retrying an operation does not bypass commit confirmation.