Cloudera Docs

Fixed Issues

Review the list of issues that are resolved in the Cloudera Observability on premises 3.7.1 release.

OBS-12229: Some columns on the HMS Tables page do not display data
Previously, on the Tables page of the Hive Metastore (HMS) service in Cloudera Observability on premises, some columns did not display any data, although the values were available. Exporting the same table data in Excel format produced a file that contained the correct values for those columns.
This issue is now resolved. All columns on the page display their values.
OBS-12232: Impala queries are not displayed in the query list after an upgrade
Previously, after you upgraded Cloudera Observability on premises from 3.5.3 to 3.7.1, Apache Impala queries that ran before the upgrade were not displayed on the Summary page or in the query list. You could retrieve such a query only by searching for its query ID, because the migrated query records were missing the time values that these pages use to filter queries by time.
This issue is resolved by calculating the missing millisecond start and end timestamps for migrated query records from the query start and end times that are already stored. Impala queries that ran before the upgrade are now displayed on the Summary page and in the query list.
OBS-12257: The top databases widget for Impala loads slowly and consumes excessive memory
Previously, the top databases widget for Apache Impala generated an inefficient query against the analytics database. The query cross-joined the matching queries with a row set derived from the entire Impala query history, and therefore the volume of data that it processed grew with the size of that history. On deployments with a large query history, the widget loaded slowly and placed severe memory pressure on the analytics database.
This issue is resolved by counting only the first 100 tables that each Impala query accesses, and by excluding empty database names earlier in the query. Almost no query accesses more than 100 tables, therefore the effect on the reported database rankings is negligible.
OBS-12571: Spark jobs are not displayed on the Summary and Engines pages
Previously, Cloudera Observability on premises could not process Apache Spark event logs on hosts that run glibc but also have musl libc libraries installed. On such hosts, the snappy-java 1.1.10.8 compression library incorrectly detected musl libc and failed to initialize its native library, which stopped the ingestion of Spark event logs. As a result, Spark jobs were not displayed on the Summary and Engines pages.
This issue is resolved. Cloudera Observability on premises now uses snappy-java version 1.1.10.7, which is not affected by the incorrect detection.
OBS-12583: The Hot Tables tab of the Data Temperature widget does not display any data
Previously, the Hot Tables tab of the Data Temperature widget did not display any data for a Hive Metastore (HMS) service, although the underlying query returned results. The query returned tables in no particular order, including tables that no query had accessed. The interface reads only the first 500 rows of the result and discards tables that have no queries, therefore the tab appeared empty whenever the queried tables were positioned after the first 500 rows.
This issue is resolved by sorting the results so that the most frequently queried tables, and then the largest tables, are returned first, and by returning only the highest ranking tables.
OBS-13491: LDAP login to the Console Server fails when the LDAP server certificate has an empty Subject field
Previously, when LDAP authentication was configured over LDAPS with a certificate authority (CA) certificate, the Cloudera Observability Console Server could not complete the initial LDAP bind if the LDAP server certificate contained no Subject field and no usable subject alternative name. Logging in with valid LDAP credentials failed with the message Internal Server Error, and the Console Server log displayed the following error: Error [ERR_TLS_CERT_ALTNAME_INVALID]: Hostname/IP does not match certificate's altnames: Cert is empty. Certificates without a Subject field are valid in some enterprise public key infrastructure (PKI) configurations.
This issue is resolved. The Console Server now skips host name verification only when the LDAP server certificate contains neither a usable common name nor a DNS or IP subject alternative name, and records a warning in the log when it does so. Validation of the certificate authority chain is always performed, and host name verification is still enforced for certificates that contain identity information.
OBS-13456: Cannot access the Cloudera Observability UI when LDAP user or group names contain non-ASCII characters
Previously, when role-based access control (RBAC) was enabled, you could not access the Cloudera Observability UI if the LDAP user name, or any LDAP group name assigned to the user, contained non-ASCII characters, such as Hebrew characters. The Console Server passes the user name and group names to the Cloudera Observability services in an HTTP header, and the header value was rejected because it contained characters outside the ASCII character set.
This issue is resolved. The Console Server now encodes the user name and group names in Base64 before sending them, and the Cloudera Observability services decode the value when they receive them.
OBS-12074: The Cloudera Observability UI displays incomplete Hive Metastore data while an extraction run is in progress
Previously, the Cloudera Observability UI displayed data from the extraction run that was in progress instead of from the last successful extraction run. When a Hive Metastore (HMS) extraction run took a long time to complete, for example longer than 24 hours, the database and table counts from the previous completed run were reset, and the UI displayed only the databases that the current run had extracted so far.
This issue is resolved. The queries that supply the UI now select the most recent successful extraction separately for each database. Databases that the current extraction run has not yet processed continue to display the data from their last successful extraction run, therefore the database and table counts remain complete while an extraction run is in progress.