Cluster management

This document mainly introduces the management tools of TuGraph HA cluster, including the functions of deleting nodes, leader transfer and generating snapshots.

1 Introduction

After the HA cluster is started, you can use the lgraph_peer tool for cluster management, which can perform functions such as deleting nodes, transferring leaders, and generating snapshots.

2. Delete node

For nodes in the TuGraph HA cluster that are offline for a long time or have network partitions, you can use the remove_peer command of lgraph_peer to delete the node. An example command is as follows:

$ lgraph_peer --command remove_peer --peer {peer_id} --conf {group_conf}

in:

--command remove_peer specifies that the operation to be performed is remove_peer, that is, delete the node.
--peer {peer_id} specifies the rpc network address of the node to be deleted, such as 127.0.0.1:9092.
--conf {group_conf} specifies the member configuration of the HA cluster (as long as it can be connected to the master node), such as 127.0.0.1:9092,127.0.0.1:9093,127.0.0.1:9094.

3. leader transfer

When you need to shut down or restart the master node, in order to reduce the unserviceable time of the cluster, you can use the transfer_leader command of lgraph_peer to transfer the master node. An example command is as follows:

$ lgraph_peer --command transfer_leader --peer {peer_id} --conf {group_conf}

in:

--command transfer_leader specifies that the operation to be performed is transfer_leader, that is, transferring the master node.
--peer {peer_id} specifies the rpc network address to become the master node, such as 127.0.0.1:9092.
--conf {group_conf} specifies the member configuration of the HA cluster (as long as it can be connected to the master node), such as 127.0.0.1:9092,127.0.0.1:9093,127.0.0.1:9094.

4. Generate snapshot

For reasons such as setting ha_snapshot_interval_s to -1 when the node starts so that snapshots are not taken by default or other reasons, when you need to tell a node generate a snapshot, you can use the snapshot command of lgraph_peer. An example command is as follows:

$ lgraph_peer --command snapshot --peer {peer_id}

in:

--command snapshot specifies that the operation to be performed is snapshot, that is, generating a snapshot.
--peer {peer_id} specifies the rpc network address of the node to generate a snapshot, such as 127.0.0.1:9092.