It has been a while since I looked at this, but I think it ended up reading from Hue groups when using the Hue interface (particularly when you used PAM as the Hue authenticator rather than LDAP, Kerberos or similar). I can't be sure, though. I didn't report this as an issue, but I believe it is working in our development system (rather than my ad-hoc initial testing cluster of three VMs and a rough LDAP/Kerberos system). ... View more

I'm getting the same behaviour with CDH 5.8.0. The only way that I've found to get past the SentryShellSolr NullPointerException is to run the command on the machine that Sentry is running on. Only those machines get the sentry-site.xml deployed to them with the required key/value pair in it.   Unfortunately, after resolving that then I instead get:   WARN security.UserGroupInformation: PriviledgedActionException as:<user-account> (auth:KERBEROS) cause:org.apache.thrift.transport.TTransportException: Peer indicated failure: Problem with callback handler Based on this post then this is actually a security issue and <user-account> isn't listed in the "sentry.service.allow.connect" setting.   Once I used the correct server and an account that was listed in sentry.service.allow.connect then "solrctl sentry ..." commands run successfully.   The only down-side now is that Solr/Sentry appears to be ignoring Linux groups as a way of identifying which role a user has, despite it working when I used a flat file and despite Solr's "sentry.provider" value being set to org.apache.sentry.provider.file.HadoopGroupResourceAuthorizationProvider.   ... View more

I've found that earlier today and followed the steps, even though the web UI doesn't make it clear that any extra steps are necessary. Also, the "add the snippets" interface appears to have changed, so it took a minute to find what I was looking for there. It breaks at point 11 of Step 5.   I'm starting from a vanilla, unmodified 5.8 Quickstart VM (Virtualbox flavour). I'd assumed it would be correctly configured to add things like Accumulo, and I'm fairly sure it defaults to packages (RPMs). OS and versions are whatever Cloudera makes available - the VM appears to be CentOS6 and I'll be installing whatever version of Accumulo the public repos have available each time I installed over the past two days (I'm connected to the Internet and downloading packages).   I've had a look in the logs, but the only mention of "csd.sh" is an error from earlier on when I first found that Accumulo service deployment wasn't creating the "accumulo" user.   The VM is provisioned with two CPUs and 16GB RAM.   Could the problem be because I'm using a single Quickstart VM and trying to install all services on one node? I'd rather not try to set up a network and multiple VMs just for some simple tests and experimentation.   Thanks. ... View more

That sounds like an odd requirement. It always seemed like all of the other services installed any necessary software before doing the configuration. In the "add service" list then Kafka explicitly says that the parcel needs to be activated or the package installed. Accumulo says nothing of the sort in the web interface.   I've installed the "accumulo" package and tried again, but it still hangs on "Initialize Accumulo" at the "csd/csd.sh [init]" step with "UNKNOWN" in the stdout and "Bad gateway" in the Role Log.   It seems that Cloudera Manager could benefit from more pre-install checks and clearer errors, rather than just hanging/failing in a non-descript way when a dependency isn't available for Accumulo (and then still doing the same when the package *is* installed).   Thanks. ... View more

This doesn't appear to work for me. I'm using the Quickstart VM (5.8.0, downloaded today) and with or without an Accumulo Gateway it hangs at " Execute command Initialize Accumulo on role Master (quickstart) ".   Running "ps aux|grep csd" doesn't show any running "csd.sh" processes. ... View more