Open
Description
Router's failover constantly pings replicas to know, if they're
alive and properly working. If replica doesn't work correctly,
then its priority is temporary lowered and callro
starts to
go to another replica. Moreover, callbro and other requests
skip this replica during balance, if it's unhealthy.
One of the reasons to mark replica as unhealthy is failed
requests in a row. This option allows to specify the number
of such failed requests to mark replica as unhealthy. Note,
that only network problems are considered as failed request.
The option is specified in the root of the config. Default value
is 3.
Requested by @Serpentian in tarantool/vshard@5570d87.
Metadata
Metadata
Assignees
Labels
No labels