PD is the abbreviation for Placement Driver. It is used to manage and schedule the TiKV cluster.
PD supports distribution and fault-tolerance by embedding etcd.
- Make sure โGoโ (version 1.8+) is installed.
- Ensure your
$GOPATH
is set. (For example,export GOPATH=$HOME/go
) - Clone the repository with
git clone [email protected]:pingcap/pd.git $GOPATH/src/github.com/pingcap/pd
. - Use
make
to install PD. PD is installed in thebin
directory.
See configuration.
You can run pd-server
directly on your local machine, if you want to connect to PD from outside,
you can let PD listen on the host IP.
# Set correct HostIP here.
export HostIP="192.168.199.105"
pd-server --name="pd" \
--data-dir="pd" \
--client-urls="http://${HostIP}:2379" \
--peer-urls="http://${HostIP}:2380" \
--log-file=pd.log
Using curl
to see PD member:
curl http://${HostIP}:2379/v2/members
{"members":[{"id":"f62e88a6e81c149","name":"pd","peerURLs":["http://192.168.199.105:2380"],"clientURLs":["http://192.168.199.105:2379"]}]}
A better tool httpie is recommended:
http http://${HostIP}:2379/v2/members
HTTP/1.1 200 OK
Content-Length: 144
Content-Type: application/json
Date: Thu, 21 Jul 2016 09:37:12 GMT
X-Etcd-Cluster-Id: 33dc747581249309
{
"members": [
{
"clientURLs": [
"http://192.168.199.105:2379"
],
"id": "f62e88a6e81c149",
"name": "pd",
"peerURLs": [
"http://192.168.199.105:2380"
]
}
]
}
You can use the following command to build a PD image directly:
docker build -t pingcap/pd .
Or you can also use following command to get PD from Docker hub:
docker pull pingcap/pd
Run a single node with Docker:
# Set correct HostIP here.
export HostIP="192.168.199.105"
docker run -d -p 2379:2379 -p 2380:2380 --name pd pingcap/pd \
--name="pd" \
--data-dir="pd" \
--client-urls="http://0.0.0.0:2379" \
--advertise-client-urls="http://${HostIP}:2379" \
--peer-urls="http://0.0.0.0:2380" \
--advertise-peer-urls="http://${HostIP}:2380" \
--log-file=pd.log
PD is a component in TiDB project, you must run it with TiDB and TiKV together, see binary deployment to learn how to set up the cluster and run them.
You can also use Docker to run the cluster.