When using encrypted passwords via the Hadoop Credential Provider, HiveServer2 currently does not correctly forward enough information to the job configuration for jobs to read those secrets. If your job needs to access any secrets, like S3 credentials, then there's no convenient and secure way to configure this today.
You could specify the decryption key in files like mapred-site.xml that HiveServer2 uses, but this would place the encryption password on local disk in plaintext, which can be a security concern.
To solve this problem, HiveServer2 should modify job configuration to include the environment variable settings needed to decrypt the passwords. Specifically, it will need to modify:
- For MR2 jobs:
- For Spark jobs:
HiveServer2 can get the decryption password from its own environment, the same way it does for its own credential provider store today.
Additionally, it can be desirable for HiveServer2 to have a separate encrypted password file than what is used by the job. HiveServer2 may have secrets that the job should not have, such as the metastore database password or the password to decrypt its private SSL certificate. It is also best practices to have separate passwords on separate files. To facilitate this, Hive will also accept:
- A configuration for a path to a credential store to use for jobs. This should already be uploaded in HDFS. (hive.server2.job.keystore.location or a better name) If this is not specified, then HS2 will simply use the value of hadoop.security.credential.provider.path.
- An environment variable for the password to decrypt the credential store (HIVE_JOB_KEYSTORE_PASSWORD or better). If this is not specified, then HS2 will simply use the standard environment variable for decrypting the Hadoop Credential Provider.