Manual Rolling Upgrade
MapR supports manual rolling upgrades from 4.0.x and higher. If you are upgrading from an earlier version, upgrade MapR core manually and offline. In a manual rolling upgrade, you upgrade the MapR software one node at a time so that the cluster as a whole remains operational throughout the process. The fileserver service on each node goes offline while packages are upgraded, but its absence is short enough that the cluster does not raise the data under-replication alarm.
The manual rolling upgrade procedure reflects the behavior of the scripted rolling upgrade. When you upgrade each node, you follow the same group upgrade order and the packages upgrade order as the script.
Group Upgrade Order
Upgrade cluster nodes in groups based on the services running on each node. Upgrade groups of nodes in the following order:
| Group | Nodes in this Group |
|---|---|
| 1 | Each node only has Zookeeper. This establishes a stable ZooKeeper quorum on the new version, which will remain active through the rest of the upgrade process. |
| 2 |
Each node only has Fileserver or it has Fileserver and Zookeeper. |
| 3 |
Each node only has TaskTracker or it has TaskTracker, FileServer, and Zookeeper. |
| 4 | Each node only has NodeManager or it has NodeManager,TaskTracker, FileServer, and Zookeeper |
| 5 | Each node only has JobTracker or it has JobTracker,NodeManager, TaskTracker, FileServer, and Zookeeper. When you upgrade nodes in this groups, upgrade nodes with the standyby JobTrackers before you upgrade the node with the active JobTracker. |
| 6 | Each node only has ResourceManager or it has ResourceManager, JobTracker, NodeManager, TaskTracker, FileServer, and Zookeeper. When you upgrade nodes in this groups, upgrade nodes with the standyby ResourceManagers before you upgrade the node with the active ResourceManager. |
| 7 |
Each node only has HBase Region server or it has HBase Region server, ResourceManager, JobTracker,NodeManager,TaskTracker, FileServer, and Zookeeper. |
| 8 | Each node only has HbaseMaster or it has HbaseMaster,Hbase Region server ,ResourceManager, JobTracker,NodeManager,TaskTracker, FileServer, and Zookeeper. |
| 9 |
Each node only has CLDB server or it has CLDB server, HBaseMaster,HBase Region server,ResourceManager, JobTracker,NodeManager, TaskTracker, FileServer, and Zookeeper. When you upgrade nodes in this group, upgrade nodes with the slave CLDB before you upgrade the node with the master CLDB. |
Package Upgrade Order
When you upgrade each node, upgrade existing packages in the following order:
- On all operating systems except SUSE, upgrade the mapr-core package first. Subsequent packages can be done in any order.
- On SUSE, upgrade the mapr-compat-suse package first and the mapr-core package second. Subsequent packages can be done in any order.
The following is a list of the primary packages:
- mapr-cldb
- mapr-compat-suse (if upgrading on SUSE)
- mapr-core-internal
- mapr-core
- mapr-fileserver
- mapr-hadoop-core
- mapr-historyserver
- mapr-jobtracker
- mapr-mapreduce1
- mapr-mapreduce2
- mapr-metrics
- mapr-nfs
- mapr-nodemanager
- mapr-resourcemanager
- mapr-tasktracker
- mapr-webserver
- mapr-zk-internal
- mapr-zookeeper
Before you Upgrade
Perform the following steps before you begin the upgrade:
-
Based on the group upgrade order, determine the upgrade groups. To see a list of services on each node, you can run the following command:
maprcli node list -columns hostname,csvc - If the cluster is secure, the MapR Admin user must have a security ticket created before running the upgrade. Otherwise, some upgrade commands will not run.