Support Questions
Find answers, ask questions, and share your expertise

How to restart (unpause a downgrade)

While attempting to downgrade a cluster,the process becomes paused. How do you reinitiate this process?


This is an Isilon backed cluster, but the logs show no issues with the component.


Is there anyway to restart/unpause a downgrade?



@russ stevenson

You may want to look into the official docs of RU

Worse case: Support ticket. "Isilon cluster "

@russ stevenson

see if it's prompting you for manual intervention, usually upgrade/downgrade will pause for user input.

@russ stevenson

Rolling upgrade is not tested on Isilon AFAIK. Please open a case with Hortonworks Support to see if you can get past the Ambari issue. Please contact the Isilon SE and get engineering help to get past this issue, and prioritize this.



We did in fact test this, and implemented changes in OneFS and to support Rolling Upgrade. See my separate answer with details.

Thanks for all responses, this is actually an internal cluster used to develop Isilon-HDP integration content for Isilon PS delivery and integration. I'm an EMC Isilon SA and was using this cluster to develop an ambari-hdp-isilon kerberos deployment methodology.

This cluster is also Kerberized which is likely contributing to the complexity, the downgrade does appear to be completely stalled out and was looking to attempting to just restart the paused downgrade.

The paused component is just reporting: The initial batch of DataNode hosts have been downgraded. You are advised to check the hosts and perform cluster/workload-specific tests against your cluster to ensure proper operation before proceeding with downgrade of the remaining services. The Isilon HDFS components all appear good and I do not appear to have the ability to take action on the downgrade tab.


If no way to trigger the downgrade to continue exists, It may be i just rebuild the cluster at this time. I have no HDP support and it's a test cluster anyway.

Thanks for all your time.


yes it is waiting for user input but I guess it has no means of doing so. It would be curious to see a way to go past it. @mahadev @smohanty @yusaku


@russ stevenson are you still having issues with this? Can you accept the best answer or post your solution? Additionally, it would really be nice to see an article about your experience with Isilon. This is very useful.


Rolling Upgrade of HDP is supported with OneFS. You'll need to be on OneFS or or higher.

Those releases implemented a new setting in OneFS, --odp-version , which allows OneFS to respond with the expected version to Ambari. Also, customers must use the experimental setting to skip NameNode HA check within Ambari. That check is not valid to the OneFS environment.

Here are the instructions:

Those instructions apply to Express Upgrade as well, except the experimental setting change is not necessary since HA is not a requirement for Express.

If you're below one of those version numbers, or if for some reason you must use Manual Upgrade, the process is messier and requires some Ambari curls or database edits. Here's the procedure if you must use Manual: