PD Recover User Guide

PD Recover is a disaster recovery tool of PD, used to recover the PD cluster which cannot start or provide services normally.

Compile from source code

  • Go Version 1.16 or later is required because the Go modules are used.
  • In the root directory of the PD project, use the make pd-recover command to compile and generate bin/pd-recover.

Download TiDB Toolkit

The PD Recover installation package is included in the TiDB Toolkit. To download the TiDB Toolkit, see Download TiDB Tools.

Quick Start

This section describes how to use PD Recover to recover a PD cluster.

Get cluster ID

The cluster ID can be obtained from the log of PD, TiKV or TiDB. To get the cluster ID, you can view the log directly on the server.

To get the cluster ID from the PD log, run the following command:

cat {{/path/to}}/pd.log | grep "init cluster id"
[2019/10/14 10:35:38.880 +00:00] [INFO] [server.go:212] ["init cluster id"] [cluster-id=6747551640615446306] ...

Get cluster ID from TiDB log

To get the cluster ID from the TiDB log, run the following command:

cat {{/path/to}}/tidb.log | grep "init cluster id"
2019/10/14 19:23:04.688 client.go:161: [info] [pd] init cluster id 6747551640615446306 ...

Get cluster ID from TiKV log

To get the cluster ID from the TiKV log, run the following command:

cat {{/path/to}}/tikv.log | grep "connect to PD cluster"
[2019/10/14 07:06:35.278 +00:00] [INFO] [tikv-server.rs:464] ["connect to PD cluster 6747551640615446306"] ...

Get allocated ID

The allocated ID value you specify must be larger than the currently largest allocated ID value. To get allocated ID, you can either get it from the monitor, or view the log directly on the server.

To get allocated ID from the monitor, you need to make sure that the metrics you are viewing are the metrics of the last PD leader, and you can get the largest allocated ID from the Current ID allocation panel in PD dashboard.

Get allocated ID from PD log

To get the allocated ID from the PD log, you need to make sure that the log you are viewing is the log of the last PD leader, and you can get the maximum allocated ID by running the following command:

cat {{/path/to}}/pd*.log | grep "idAllocator allocates a new id" | awk -F'=' '{print $2}' | awk -F']' '{print $1}' | sort -r -n | head -n 1
4000 ...

Or you can simply run the above command in all PD servers to find the largest one.

Deploy a new PD cluster

Before deploying a new PD cluster, you need to stop the the existing PD cluster and then delete the previous data directory or specify a new data directory using --data-dir.

Use pd-recover

You only need to run pd-recover on one PD node.

./pd-recover -endpoints http://10.0.1.13:2379 -cluster-id 6747551640615446306 -alloc-id 10000

Restart the whole cluster

When you see the prompted information that the recovery is successful, restart the whole cluster.

FAQ

Multiple cluster IDs are found when getting the cluster ID

When a PD cluster is created, a new cluster ID is generated. You can determine the cluster ID of the old cluster by viewing the log.

The error dial tcp 10.0.1.13:2379: connect: connection refused is returned when executing pd-recover

The PD service is required when you execute pd-recover. Deploy and start the PD cluster before you use PD Recover.