Member since
09-04-2015
6
Posts
0
Kudos Received
0
Solutions
03-13-2023
08:10 AM
Thank you! We've engaged Dell but it's been .... slow 🙂 I'll ask my customer to reach out to Cloudera Support as well, that's a good idea.
... View more
03-13-2023
01:25 AM
We're trying to install CDP 7.1.7 on Isilon with OneFS 8.2.2 We found several documents that we tried to follow: - https://infohub.delltechnologies.com/section-assets/h18730-dell-emc-powerscale-onefs-cdp-private-base-install-guide - https://dl.dell.com/content/docu92689_PowerScale_OneFS_HDFS_Reference_Guide.pdf?language=en_US&source=Coveo - https://infohub.delltechnologies.com/l/powerscale-onefs-user-mapping-mapping-identities-across-authentication-providers/mapping-rules-cannot-contain-user-principal-name And more.... We also used the isilon hadoop tools and found a few bugs which we provided Pull Requests for (https://github.com/Isilon/isilon_hadoop_tools/pull/105 & https://github.com/Isilon/isilon_hadoop_tools/pull/106 ) so it seems as if thas hasn't been used in a while. We were able to finish the basic setup but we do have issues creating the proper User Mapping Rules. The docs say: "You cannot use a user principal name in a user mapping rule. A user principal name (UPN) is an Active Directory domain and username that are combined into an Internet-style name with an @ sign, like an email address: jane@example.com. If you include a UPN in a rule, the mapping service ignores it and might return an error." Unfortunately, something doesn't quite work. Cloudera Manager automatically created all users in Active Directory for us which means that an impala user could end up having the name "cloudera_xnTfsrendtr" in Active Directory. It does have the correct UPN set ("impala@...") but the UPN can't be used in a mapping. This means for us that we have to create User Mapping Rules for all users manually which "join" the identities and maps one of those auto-generated users to real ones "cloudera_xnTfsrendtr" -> impala. That is very cumbersome and the CDP / Isilon install docs don't mention this at all. Are there updated docs anywhere/does anyone have notes on this/done this recently?
... View more
Labels:
- Labels:
-
Cloudera Data Platform (CDP)
-
Kerberos
09-15-2015
11:47 PM
Hi aBBy007, hmm I'm not sure what's happened to you but the method I described definitely works. I've done it tens of times as lots of customers don't allow direct Internet access to their clusters. Cheers, Lars
... View more
09-04-2015
08:26 AM
Not that I'm aware of. If you want to keep all your data you also need to move the data directories for some of the components to the new server.
... View more
09-04-2015
08:24 AM
That only installs CM and it's using the binary installer which uses embedded databases etc. which is not recommended for production. It also does not install CDH on cluster machines. CM itself is easy to install silently using DEBs or RPMs. To deploy CDH itself using CM would require you to use the CM REST API. I believe this is how Cloudera Director works as well because that basically does exactly what you're asking for. http://cloudera.github.io/cm_api/apidocs/v10/index.html I've seen people use the API in Ansible/Chef/Puppet scripts.
... View more
09-04-2015
12:58 AM
I don't think there's a way to disable the URL checking entirely. But you can just delete all but one Repository in the Parcel settings and enter a bogus URL (i.e. http://foo.bar) in the last field. You'll get warnings in the CM Server logs but those are quite common anyway because lots of customers don't have direct Internet access to their CM machines. And then you can just download the parcels and .sha1 files and move them into the parcel-repo directory. But you will have to rename the .sha1 file to .sha! This works but is not the recommended way of doing it. The recommended way is to set up a simple HTTP server and mirror the Parcels you want pointing the HTTP server at it and then enter that URL in CM.
... View more