That is correct. When you bring down a RM the other two will keep the cluster up and running and all the NMs will keep working with other two while the third one is being upgrade.
On 12/18/2017 09:17 AM, Debraj Manna wrote: > Thanks AD. > > When you say one at a time it means I should upgrade a ResourceManager node > and once it is fully upgraded and up, move to the next ResourceManager node? > > On Mon, Dec 18, 2017 at 7:41 PM, A D <[email protected]> wrote: > >> I strongly recommend following >> http://hadoop.apache.org/docs/r2.6.5/hadoop-project-dist/hadoop-hdfs/ >> HdfsRollingUpgrade.html >> for rolling upgrade. >> >> HBase is no different when it comes to rolling upgrade. There is no >> special command that you need to execute. Below are the steps that I >> would follow: >> >> 1. Upgrade the HDFS: >> >> Once HDFS upgrade is complete: >> >> 2. Upgrade ResourceManager nodes (one at a time). >> 3. Upgrade NodeManagers(one at a time). >> >> Follow step 2 and 3 for HBase master and RegionServers. I would wait for >> RIT to settle after shutting down a region server. >> >> Please refer to product documentation in case you are using a particular >> distribution (Cloudera, MapR or Hortonworks). >> >> >> On 12/18/2017 08:52 AM, Debraj Manna wrote: >>> I am having a cluster like below >>> >>> >>> - 3 Hbase Master (1 Active & 2 standby) >>> - 4 Region Servers >>> - 4 Data Nodes >>> - 1 Primary & 1 Secondary Name Node >>> - 3 Journal Node >>> - 4 Nodemanager >>> - 3 Resource Manager (1 Active & 2 standby) >>> >>> >>> Can someone let me know how can I plan a rolling upgrade of the above >>> cluster with minimum downtime? >>> >>> >>> - Hadoop - 2.6.0 >>> - Yarn - 2.6.0 >>> - Hbase - 1.2.0 >>> >>> Thanks, >>> >> -- >> Thank you, >> A.D >> [email protected] >> >> -- Thank you, A.D [email protected]
