A mapping fails because the Model Repository Service cannot communicate with the Data Integration Service. This happens when the following conditions are true:

The Model Repository Service is configured with an SSL-enabled database
The JDBC connection string and secured JDBC parameter are configured separately on the Model Repository Service

Additionally, the Data Integration Service does not successfully restart if it is enabled for big data recovery.

Workaround: Configure both the JDBC connection and the secured JDBC parameter in the JDBC connection string. For example:

jdbc:informatica:db2://hostname.company.com:60012;DatabaseName=DB1a2b3c;EncryptionMethod=SSL;HostNameInCertificate=hostname.company.com;ValidateServerCertificate=true;TrustStore=/home/toolprod/INFA_Automation/newssldb/DB2_SSL_CERT;TrustStorePassword=informatica

"<file> is not a Parquet file. expected magic number at tail"

A mapping with flat file sources and targets that uses the Spark engine to run on a WANdisco-enabled Hortonworks HDP 2.6.5 cluster fails.

Workaround:

Copy the following .jar files from the cluster in the
usr/hdp/<version>/hadoop/client
directory to each node where the Informatica Data Integration Service is installed to the
<Informatica home>/services/shared/spark/lib_spark_<version>
directory:
hadoop-common-x.x.x.jar
hadoop-auth-x.x.x.jar
where "x.x.x" is the .jar file version that WANdisco uses.
For example, find the file hadoop-common-2.7.3.2.6.5.0-292.jar in the
/usr/hdp/2.6.5.0-292/hadoop/client
directory on the cluster.
Restart the Data Integration Service on each node.
On the cluster node where HDFS sources/targets reside, grant 733 permisson to HDFS.

If a Hive target table is partitioned by a Date data type, a partition in the Hive target table cannot be truncated on a Hortonworks HDP 3.1 cluster.

Workaround: Add the following property to the end of the Data Access Connection String: ?hive.optimize.sort.dynamic.partition=true