tarantool / tarantool / 28510487066

88%

push

github

config: add sharding.rebalancer_bucket_send_timeout

Expose vshard's rebalancer_bucket_send_timeout option in the  config
under the sharding section. It is forwarded to vshard like the other
sharding options and is annotated with vshard_since('0.1.41'), so
configuring it with an older vshard is rejected at the configuration
validation stage.

It is a prerequisite for the recovery point manager, whose per-attempt
timeout must exceed the rebalancer bucket send timeout to guarantee that a
recovery point can be created during rebalancing.

@TarantoolBot document
Title: New sharding.rebalancer_bucket_send_timeout configuration option
Since: VShard 0.1.41 and Tarantool 3.8.0

The `sharding.rebalancer_bucket_send_timeout` option sets the timeout
(in seconds) for sending a single bucket during rebalancing. It is
global-scoped. Infinity by default.

Note: Can be used to limit the time, bucket can be unavailable for write
during bucket sending. Can be also used with `router.map_callrw` request
and with `recovery-point-manager` role (which uses the `map_callrw`
under the hood) to guarantee, that the request will succeed. The timeout
for request or role should be > 1.5 * `rebalancer_bucket_send_timeout`.

70867 of 119070 branches covered (59.52%)

106436 of 121354 relevant lines covered (87.71%)

1369278.91 hits per line