The state of a Group Replication group member shows its current
role in the group. The Performance Schema table
replication_group_members shows the
state for each member in a group. If the group is fully functional
and all members are communicating properly, all members report the
same state for all other members. However, a member that has left
the group or is part of a network partition cannot report accurate
information on the other servers. In this situation, the member
does not attempt to guess the status of the other servers, and
instead reports them as unreachable.
A group member can be in the following states:
The server is an active member of a group and in a fully functioning state. Other group members can connect to it, as can clients if applicable. A member is only fully synchronized with the group, and participating in it, when it is in the
The server has joined a group and is in the process of becoming an active member. Distributed recovery is currently taking place, where the member is receiving state transfer from a donor using a remote cloning operation or the donor's binary log. For more information, see Section 18.5.3, “Distributed Recovery”.
The Group Replication plugin is loaded but the member does not belong to any group.
The member is in an error state and is not functioning correctly as a group member. A member can enter error state either while applying transactions or during the recovery phase. A member in this state does not participate in the group's transactions. For more information on possible reasons for error states, see Section 18.7.6, “Responses to Failure Detection and Network Partitioning”.
Depending on the exit action set by
group_replication_exit_state_action, the member is in read-only mode (
super_read_only=ON) and could also be in offline mode (
offline_mode=ON). Note that a server in offline mode following the
OFFLINE_MODEexit action is displayed with
OFFLINE. A server with the exit action
ABORT_SERVERshuts down and is removed from the view of the group. For more information, see Section 184.108.40.206, “Exit Action”.
The local failure detector suspects that the member cannot be contacted, because the group's messages are timing out. This can happen if a member is disconnected involuntarily, for example. If you see this status for other servers, it can also mean that the member where you query this table is part of a partition, where a subset of the group's servers can contact each other but cannot contact the other servers in the group. For more information, see Section 18.5.4, “Network Partitioning”.
See Section 18.4.3, “The replication_group_members Table” for an example of the Performance Schema table contents.