Atlas 0.8's Metadata Server Wouldn't Start on HDP 2.6.0

william_lu1 — Thu, 30 Nov 2017 16:01:12 GMT

I just installed Atlas 0.8 on an HDP 2.6.0 cluster, which was recently upgraded from HDP 2.3.2 and Kerberized. Yet Atlas 0.8's Metadata Server wouldn't start, with these error messages:

KeeperErrorCode = NoNode for /infra-solr/collections/vertex_index org.apache.zookeeper.KeeperException$NoNodeException: KeeperErrorCode = NoNode for /infra-solr/collections/vertex_index at org.apache.zookeeper.KeeperException.create(KeeperException.java:111) at org.apache.zookeeper.KeeperException.create(KeeperException.java:51) at org.apache.zookeeper.ZooKeeper.getACL(ZooKeeper.java:1330) at org.apache.ambari.logsearch.solr.util.AclUtils.setRecursivelyOn(AclUtils.java:76) at org.apache.ambari.logsearch.solr.util.AclUtils.setRecursivelyOn(AclUtils.java:70) at org.apache.ambari.logsearch.solr.commands.SecureZNodeZkCommand.executeZkCommand(SecureZNodeZkCommand.java:45) at org.apache.ambari.logsearch.solr.commands.SecureZNodeZkCommand.executeZkCommand(SecureZNodeZkCommand.java:32) at org.apache.ambari.logsearch.solr.commands.AbstractZookeeperRetryCommand.createAndProcessRequest(AbstractZookeeperRetryCommand.java:38) at org.apache.ambari.logsearch.solr.commands.AbstractRetryCommand.retry(AbstractRetryCommand.java:45) at org.apache.ambari.logsearch.solr.commands.AbstractRetryCommand.run(AbstractRetryCommand.java:40) at org.apache.ambari.logsearch.solr.AmbariSolrCloudClient.secureZnode(AmbariSolrCloudClient.java:177) at org.apache.ambari.logsearch.solr.AmbariSolrCloudCLI.main(AmbariSolrCloudCLI.java:522)

The zookeeper-client indicates that the znode /infra-solr/configs/atlas_configs does exist, yet none of the Atlas's intended collections (namely, vertex_index, edge_index, and fulltext_index) were created.

I attempted to delete the znode /infra-solr/configs/atlas_configs and start from scratch, but I was blocked by an authentication error, as shown below:

rmr /infra-solr/configs/atlas_configs

Authentication is not valid : /infra-solr/configs/atlas_configs/solrconfig.xml

So, how am I going to get out of this?

For reference, here is the sequence of operations performed on the cluster:

<1> Starting Point: HDP 2.3.2 managed by Ambari Server 2.1.2.

<2> Upgraded Ambari Server to 2.5.2.0.

<3> Installed Ambari Infra (including Solr).

<4> Deleted Atlas 0.5.0.

<5> Implemented Kerberos.

<6> Upgraded to HDP 2.6.0.

<7> Installed Atlas 0.8.0.

Now, is it possible that the znode /infra-solr/configs/atlas_configs was actually created by Atlas 0.5.0 when the cluster was still on HDP 2.3.2? (Which is probably why I cannot authenticate and delete it?)

Many thanks!

Re: Atlas 0.8's Metadata Server Wouldn't Start on HDP 2.6.0

william_lu1 — Tue, 12 Dec 2017 06:24:41 GMT

In fact, this is the combined effect of two issues:

<1> The atlas.war file was not extracted properly during the Atlas installation. So, manually do so, as detailed here:

https://community.hortonworks.com/questions/104751/index.html

<2> The bug AMBARI-18368 caused a number of Atlas configuration files to be absent from their intended locations. The instructions to address this issue is here:

https://issues.apache.org/jira/browse/AMBARI-18368

Now, Atlas is up, finally.

question Re: Atlas 0.8's Metadata Server Wouldn't Start on HDP 2.6.0 in Archives of Support Questions (Read Only)

Atlas 0.8's Metadata Server Wouldn't Start on HDP 2.6.0

Re: Atlas 0.8's Metadata Server Wouldn't Start on HDP 2.6.0