CDH Community Edition Rolling Upgrade from 5.7 to 5.13

狂风中的少年 提交于 2019-12-13 17:42:58

问题


Can someone let me know how can I perform Rolling Upgrade of CDH from 5.7 to 5.13? I could not find much in Cloudera documentation regarding rolling upgrade of CDH Community Edition?

EDIT

As per the discussion below I can do upgrades manually stopping, upgrading (via "1 Click Install" )and starting the nodes. In a cluster like below

  1. 3 Hbase Master (1 Active & 2 standby)
  2. 4 Region Servers
  3. 4 Data Nodes
  4. 1 Primary & 1 Secondary Name Node
  5. 3 Journal Node
  6. 4 Nodemanager
  7. 3 Resource Manager (1 Active & 2 standby)

What should be the order of stopping, upgrading & starting of components? Is it like below:-

  1. Stop, Upgrade & Start all data nodes one by one.
  2. Stop, Upgrade & Start region servers one by one
  3. Stop, Upgrade & Start Upgrade HBase standby nodes one by one
  4. Stop, Upgrade & Start Secondary Name Node
  5. Upgrade Primary Name Node
  6. Upgrade Journal Nodes
  7. Upgrade HBase Master
  8. Stop, upgrade & Start standby Resource Manager one by one
  9. Stop, upgrade & Start Active Resource Manager
  10. Stop, Upgrade & Start Nodemanagers one by one.

If I don't follow any order and update the services in the nodes in parallel, will I face any issue (as this is just a minor version upgrade)?

If I am taking the entire cluster down by stopping all the services in all the nodes and upgrading and then starting? What order should I follow while starting the services?


回答1:


I think you have to do it manually since you are using Community Version.

A rolling upgrade allows you to upgrade your cluster software and restart the upgraded services without taking the cluster offline. Performing a rolling upgrade requires the following:

  • CDH was installed using Cloudera Manager and parcels. You can migrate your cluster from using packages to using parcels.
  • The cluster uses a Cloudera Enterprise license.

The upgrade overview page has the information you need, but it generally follows

  1. Backup Cloudera Manager database
  2. Upgrade Cloudera Manager server
  3. Upgrade Cloudera Manager Agents on all hosts
  4. Upgrade CDH parcels
  5. Reboot cluster. Try to do it per service. For example, make sure all Cloudera agents are healthly before starting Hadoop services. Then start the namenode(s), and datanodes, followed by resourcemanager(s), etc.

If you are not using Cloudera Manager, you will have to go through the command line upgrade procedures.

  • Before Upgrading to the Latest Release of CDH
  • Upgrade CDH to latest

From those you listed, you would stop all services before doing anything. For example, HBase isn't going to work if the datanodes and namenodes are down.

Build a dependency tree, and shut it down from the most dependent services. Meaning, HBase off, then YARN (one resource manager at a time), then HDFS (datanodes, then namenodes and journalnodes). Then upgrade and start them in reverse order.


The second link there says this order for upgrades

  • NameNode into safemode
  • Stop all Hadoop services
  • Backup the namenode
  • Download the 1-click packages
  • Upgrade YARN (which requires upgrading Zookeeper)
  • Upgrade JournalNodes
  • Start HDFS Service
  • Start YARN
  • Upgrade rest of components


来源:https://stackoverflow.com/questions/47660125/cdh-community-edition-rolling-upgrade-from-5-7-to-5-13

易学教程内所有资源均来自网络或用户发布的内容,如有违反法律规定的内容欢迎反馈
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!