Setting up an external Zookeeper Solr Cluster
This is a step by step instruction on how to create a cluster that has three Solr nodes running in cloud mode. These instructions should work on both a local cluster (for testing) and a remote cluster where each server runs in its own physical machine.
This was tested on
Solr version 5.4.1 and
Zookeeper version 3.4.6
Installing Solr and Zookeeper
- Download and extract Solr:
curl -O http://archive.apache.org/dist/lucene/solr/5.5.3/solr-5.5.3.tgz
tar -zxvf solr-5.4.1.tgz -C /opt/solr --strip-components=1
- Download and extract ZooKeeper:
curl -O http://archive.apache.org/dist/zookeeper/zookeeper-3.4.6/zookeeper-3.4.6.tar.gz
tar -zxvf zookeeper-3.4.6.tar.gz -C /opt/zookeeper --strip-components=1
Configuring a ZooKeeper quorum
Create our data directories:
mkdir -p /data/zookeeper/z1/data
mkdir -p /data/zookeeper/z2/data
mkdir -p /data/zookeeper/z3/data
Add our server ids at each data directory:
echo 1 > /data/zookeeper/z1/data/myid echo 2 > /data/zookeeper/z2/data/myid echo 3 > /data/zookeeper/z3/data/myid
Copy the sample config file into the first data directory:
cp ./conf/zoo_sample.cfg /data/zookeeper/z1/zoo.cfg
Open /data/zookeeper/z1/zoo.cfg in your text editor and update/add the following values:
dataDir=/data/zookeeper/z1/data clientPort=2181 # Our zookeeper quorum: # Note: If you want to run each server in its own machine, # change the ip address in each entry; Obviously, you # do not have to use a different port number for each server. # each entry is defined as follows: # server.<value_of_myid_file_in_data_dir>: server.1=127.0.0.1:2222:2223 server.2=127.0.0.1:3333:3334 server.3=127.0.0.1:4444:4445
We configured the
dataDir, which tells Zookeeper where to store its data, and
clientPort, which assigns the port on which a client (Solr in this case) should use to connect to Zookeeper.
Lastly, we defined the entries of our zookeeper quorum. Each entry is defined as server.id=host:port:port, where id is the server id number that we previously defined in the
myid file and the host is for the IP address of the Zookeeper instance to communicate with other Zookeeper instances. The port has two parts. The first is the port that followers use to connect to the leader, and the second is for leader election. For more information, refer to Zookeeper's configuration parameters documentation.
Copy the config file into the data directories of the other two servers:
cp /data/zookeeper/z1/zoo.cfg /data/zookeeper/z2/ cp /data/zookeeper/z1/zoo.cfg /data/zookeeper/z3/
[This step is ONLY required when running the servers on the same host] Change both the dataDir and clientPort for the last two servers:
- Server 2:
$ vi /data/zookeeper/z2/zoo.cfg dataDir=/data/code/zookeeper/z2/data clientPort=2182
- Server 3:
$ vi /data/zookeeper/z3/zoo.cfg dataDir=/data/code/zookeeper/z3/data clientPort=2183
Start the servers:
$ ./bin/zkServer.sh start /data/zookeeper/z1/zoo.cfg $ ./bin/zkServer.sh start /data/zookeeper/z2/zoo.cfg $ ./bin/zkServer.sh start /data/zookeeper/z3/zoo.cfg
Ensure that there are no major errors in the log file:
$ cat zookeeper.out
- Start three Solr instances and have them point at our Zookeeper instances:
# If you are running the Zookeeper servers on remote machines, use # the IP address of each server instead of the localhost. $ ./bin/solr start -c -p 8983 -z 127.0.0.1:2181,127.0.0.1:2182,127.0.0.1:2183 $ ./bin/solr start -c -p 8984 -z 127.0.0.1:2181,127.0.0.1:2182,127.0.0.1:2183 $ ./bin/solr start -c -p 8985 -z 127.0.0.1:2181,127.0.0.1:2182,127.0.0.1:2183
- Upload our collection configuration to ZooKeeper:
$ ./server/scripts/cloud-scripts/zkcli.sh -cmd upconfig \ -zkhost 127.0.0.1:2181 \ -confdir ./server/solr/configsets/data_driven_schema_configs/conf/ \ -confname my-config
Create a Solr collection using the uploaded configuration.
- If you want to create multiple collections with different schemas, then repeat the last two steps for each collection that uses a different schema. Otherwise, Zookeeper will sync the schema for all collections and you will end up with a single schema for all collections.
- In Solr, the default maxShardsPerNode is one shard per node. In this setup, we had 3 nodes, so we should not attempt to add more replicas to a collection (e.g., numShards=2 & replicationFactor=2 will result in four shards in total spreaded across three nodes). This would cause a series of errors and crashes since two replicas of the same shard will never be allowed to exist on the same node as per the maxShardsPerNode config setting.