11-12-2017 03:15 PM - last edited on 11-13-2017 05:38 AM by cjervis
I'm attempting to upgrade our CDH cluster from v5.12.1 to v5.13.0. I've succesffully updated CM and all agents to v5.13.0 in preparation for the upgrade. I've added an additional CDH parcel URL, and checked for new parcels but have thus far had no luck getting the Cloudera console to show new parcels available for CDH 5.13.0.
I've tried the following parcel URL's so far with no luck:
Am I missing something here?
11-12-2017 04:42 PM - edited 11-12-2017 04:51 PM
This cluster was deployed using Cloudera Director client (bootstrap-remote)
On the Cloudera manager node, I see the following errors in cloudera-scm-server.log
2017-11-13 00:49:57,323 INFO 742270861@scm-web-119:com.cloudera.server.web.cmf.ParcelController: Synchronizing repos based on user request admin
2017-11-13 00:49:58,080 WARN ParcelUpdateService:com.cloudera.cmf.persist.ReadWriteDatabaseTaskCallable: Error while executing CmfEntityManager task
I've seen similar threads attribute parcel problems to proxy settings, however I don't have problems adding parcels for other products; i.e. Kudu, Kafka.
11-13-2017 09:54 AM
We started seeing the same exception in Cloudera Manager on all our Cloudera clusters recently (5.8 - 5.12).
I figured the problematic parcel came from https://archive.cloudera.com/sqoop-connectors/parcels/latest/, which was included in Parcels Configuration by default. Excluding that URL from the list solved the problem.
11-14-2017 11:03 AM - edited 11-14-2017 11:14 AM
Thanks for the reply.
What was problematic about this URL? I see it in our list of parcel URLs, with an "Available Remotely" status.
I do see the following parcel errors:
Error for parcel ACCUMULO-1.4.4-1.cdh4.5.0.p0.65-el7 : Parcel not available for OS Distribution RHEL7.
Error for parcel SPARK-0.9.0-1.cdh4.6.0.p0.98-el7 : Parcel not available for OS Distribution RHEL7.
I've removed these from the parcel config, but they still show up in the parcels list. Did you need to restart your agents?
11-14-2017 11:18 AM
You do not need to restart your agents.
Even if you remove the URL from the parcels config page they will show up if they are active or even downloaded on ANY hosts in any cluster CM manages.
Say you have Cluster 2 hosts running RHEL6 and those parcels were downloaded to those hosts, they will show on the cluster parcels page.
If this is not the case with your cluster, it is hard to trouble shoot remotely. Sorry.
11-14-2017 11:18 AM
I removed the sqoop repo and that did the trick! The parcel errors for accumulo and spark cleared out as well. Upon checking for new parcels I no longer see the java exceptions in the cloudera manager logs.
11-20-2017 04:54 PM
We recently hit this issue also, where our build automation failed because no parcels were initially available to CM. Someone from Cloudera may want to doublecheck http://archive.cloudera.com/sqoop-connectors/parcels/latest/manifest.json, which appears to be invalid compared to http://archive.cloudera.com/sqoop-connectors/parcels/1.6/manifest.json
11-21-2017 06:28 AM
This issue is being looked at. Sounds like same issue as http://community.cloudera.com/t5/Cloudera-Manager-Installation/Can-t-find-cloudera-5-7-parcel/td-p/6...