Hello @pablobhz , Please look at the ODBC driver documentation, installation guide (this is for version 2.8.0): https://www.cloudera.com/downloads/connectors/hive/odbc/2-8-0.html Which driver version are you using? And what is the cluster version? The "AuthMech=1" corresponds to the Kerberos based authentication mode, that's why the connector tried to connect with Kerberos. The "AuthMech=0" means no authentication is needed. Is the cluster Kerberos enabled? If yes, then it is not possible to connect without some credentials. If the cluster (Hive service) is indeed kerberos enabled, then the "No Kerberos credentials available" and the "default cache: KEYRING:persistent:0" shows that the krb5.conf / "default_ccache_name " has been configured to use keyring. Instead of that please use a ticket cache file with the "FILE" based, see https://web.mit.edu/kerberos/krb5-1.12/doc/basic/ccache_def.html  As I see there is no "KerberosAuth" configuration option for the latest ODBC driver.  Thanks  Miklos ... View more

Right, thanks for verifying that the partition locations are not renamed. Then it is clearly the HIVE-26158, which has been fixed in 7.1.7 SP2, please plan/test and apply the 7.1.7 SP2 or the even newer 7.1.7 SP3 version (recommended) to avoid this in the future. (one should not need to alter the partition locations when table is renamed) Best regards  Miklos ... View more

Yes, with renaming external tables Hive does not change their locations (to the new desired path, irrespective whether the DB changes or the table name), so once you drop the renamed table, due to the 'external.table.purge'='true' it dropped the original location's data directory. FYI there were some fixes around the rename functionality, HIVE-24920 fixed in 7.1.7 SP1 and HIVE-26158 fixed in 7.1.7 SP2, not sure if that affects you but it's recommended to apply the latest service pack. ... View more

Hi @Patriciabqc  The behavior depends on what was the table type and it's table properties. The CDP version may be also important as there were some fixes around the rename functionality (whether HDFS location is changed) between the versions. The behavior is different for managed and external tables. If you rename a managed table, it should rename the HDFS location too, so you should not end up in such situations. Because of that I assume these were external tables. The table property 'TRANSLATED_TO_EXTERNAL'='TRUE' should make an external table behave like a managed table ("legacy CDH managed table" behavior), so when you rename the table the location is also changed, so probably we can rule out this too, but I mention this for clarity that this is also an option - and maybe a way to avoid this in the future. Last but not least, for external tables if the table property 'external.table.purge'='true' is there, then it will delete the HDFS location when the table is dropped. The symptoms suggest this was the case. I hope this explains. Best regards, Miklos ... View more

Hi @ipson , please see the CML docs: https://docs.cloudera.com/machine-learning/1.5.4/import-data/topics/ml-access-cdw-from-cml.html  which suggests to use Impyla: https://github.com/cloudera/impyla  This does not require the ODBC driver to be installed. Cheers, Miklos ... View more

Hi @szhao , based on the symptoms this may be a problem within the ODBC driver. Can you please open a support ticket through the MyCloudera support portal so that our support team can forward these information to our driver team? Thanks, Miklos ... View more

Most likely. Try to figure out the differences between those VMs. The SSPI is the Windows' integrated Kerberos library, it also has tools for "kinit" and "klist" but it's kerberos configuration is in a different place and works only if the VM is part of the Active Directory domain and the user is logged in with it's AD username. ... View more

Thanks for checking. That error message "The specified target is unknown or unreachable" is not too descriptive, but suggests some Kerberos related setup problem. If you are connecting to a CDP 7.x cluster, please use the latest Cloudera Impala ODBC 2.7.x driver as it may contain additional fixes: https://www.cloudera.com/downloads/connectors/impala/odbc/ Also you can try to add another environment variable KRB5_CONFIG and point to the krb5.ini file used by the MIT kerberos client. (c:\ProgramData\MIT\Kerberos5\krb5.ini) If none of these help, please open a support case through our support portal where you can attach more details (and the driver logs). Thank you  Miklos ... View more

Hi @pnmateus , There are a couple of points to check: - Have you created a "System wide" datasource or a "user specific" datasource in the "ODBC Data sources"?  - Per the screenshot you are using a 64 bit driver. That works only for 64 bit applications. Please verify if your Power BI (and the started python process) is also 64 bit (use Windows Task Manager to determine that) - When using MIT kerberos, the best is to define a User specific environment variable (go to your system settings / Environment variables) with name "KRB5CCNAME" and set it to a path / file which can be read only by your user. This ensures that MIT Kerberos tool will create the kerberos ticket cache file in that location and all the processes started by your user will also use the same ticket cache file when authenticating to the services (like Impala through ODBC). - If that still does not help, ultimately collect the ODBC driver logs, enable the DEBUG or Trace level logging in the datasource settings / Logging options. ... View more

Hi @shrikantverma  This indeed seems to be a generated query, however based on this only it is not clear if this is coming from the Hive Metastore. You can review with your DB tools which user is submitting these queries, for example for MySQL use "show processlist" or slow query logging to capture which user is submitting these. Alternatively stop the Hive metastore (if allowed) and observe if those queries are still appearing. As I see this query might be used by HMS while listing partitions by name in MetaStoreDirectSql.java#getPartitionsUsingProjectionAndFilterSpec(). Is this inefficient in your database? Have you checked that with an "explain <query>" or why do you want to optimize this? The HMS database should have indexes on the affected tables (DBS, TBLS, PARTITIONS) to speed up these queries, the only scan needed is to match the partition names - which can be slower if the Hive table has too many partitions. Best regards  Miklos ... View more