health_error is the last error we got from health check, or empty is the server is healthy. This is used for subset selection, we do not send queries to servers that are not healthy.
seconds_behind_master is populated for slaves only. It indicates how far behind on (MySQL) replication a slave currently is. It is used by clients for subset selection (so we don't try to send traffic to tablets that are too far behind). NOTE: This field must not be evaluated if "health_error" is not empty. TODO(mberlin): Let's switch it to int64 instead?
bin_log_players_count is the number of currently running binlog players. if the value is 0, it means that filtered replication is currently not running on the tablet. If >0, filtered replication is running. NOTE: This field must not be evaluated if "health_error" is not empty.
seconds_behind_master_filtered_replication is populated for the receiving master of an ongoing filtered replication only. It specifies how far the receiving master lags behind the sending master. NOTE: This field must not be evaluated if "health_error" is not empty. NOTE: This field must not be evaluated if "bin_log_players_count" is 0.
cpu_usage is used for load-based balancing
qps is the average QPS (queries per second) rate in the last XX seconds where XX is usually 60 (See query_service_stats.go).
bin_log_players_count is the number of currently running binlog players.
bin_log_players_count is the number of currently running binlog players. if the value is 0, it means that filtered replication is currently not running on the tablet. If >0, filtered replication is running. NOTE: This field must not be evaluated if "health_error" is not empty.
cpu_usage is used for load-based balancing
health_error is the last error we got from health check, or empty is the server is healthy.
health_error is the last error we got from health check, or empty is the server is healthy. This is used for subset selection, we do not send queries to servers that are not healthy.
qps is the average QPS (queries per second) rate in the last XX seconds where XX is usually 60 (See query_service_stats.go).
seconds_behind_master is populated for slaves only.
seconds_behind_master is populated for slaves only. It indicates how far behind on (MySQL) replication a slave currently is. It is used by clients for subset selection (so we don't try to send traffic to tablets that are too far behind). NOTE: This field must not be evaluated if "health_error" is not empty. TODO(mberlin): Let's switch it to int64 instead?
seconds_behind_master_filtered_replication is populated for the receiving master of an ongoing filtered replication only.
seconds_behind_master_filtered_replication is populated for the receiving master of an ongoing filtered replication only. It specifies how far the receiving master lags behind the sending master. NOTE: This field must not be evaluated if "health_error" is not empty. NOTE: This field must not be evaluated if "bin_log_players_count" is 0.
(Since version 0.6.0) Use toPMessage
(Since version 0.6.0) Use getField that accepts a ScalaPB descriptor and returns PValue
RealtimeStats contains information about the tablet status
health_error is the last error we got from health check, or empty is the server is healthy. This is used for subset selection, we do not send queries to servers that are not healthy.
seconds_behind_master is populated for slaves only. It indicates how far behind on (MySQL) replication a slave currently is. It is used by clients for subset selection (so we don't try to send traffic to tablets that are too far behind). NOTE: This field must not be evaluated if "health_error" is not empty. TODO(mberlin): Let's switch it to int64 instead?
bin_log_players_count is the number of currently running binlog players. if the value is 0, it means that filtered replication is currently not running on the tablet. If >0, filtered replication is running. NOTE: This field must not be evaluated if "health_error" is not empty.
seconds_behind_master_filtered_replication is populated for the receiving master of an ongoing filtered replication only. It specifies how far the receiving master lags behind the sending master. NOTE: This field must not be evaluated if "health_error" is not empty. NOTE: This field must not be evaluated if "bin_log_players_count" is 0.
cpu_usage is used for load-based balancing
qps is the average QPS (queries per second) rate in the last XX seconds where XX is usually 60 (See query_service_stats.go).