Known Issues and Limitations
Lists known issues and limitations in HPE Ezmeral Data Fabric Object Store.
HPE Ezmeral Data Fabric Object Store
has the following known issues and limitations.
- You can only test account policies through the CLI
- The account administrator can create and edit an account policy from the Object
Store UI and CLI (
/opt/mapr/bin/mc
). The account policy can only be tested through the CLI. - Any user granted permission to perform operations in an account must perform the operations from the CLI
- An account administrator can perform operations in an account from the Object Store
UI or CLI (
/opt/mapr/bin/mc
), such as creating users, groups, policies, and buckets. However, if the account administrator grants another user permission to perform operations, that user can only perform operations from the CLI (/opt/mapr/bin/mc
). The user cannot perform operations from the Object Store UI. - You cannot create buckets in non-default accounts using awscli, s3cmd, SDK
- Currently, you cannot use awscli, s3cmd, or SDK to create buckets in an account. You
must create buckets from the Object Store UI or the
/opt/mapr/bin/mc
command. After you create a bucket, you can perform all operations through any interface, including awscli, s3cmd, SDK, Object Store UI, and/opt/mapr/bin/mc
. This behavior does not apply to the default account. If you have permissions and keys (accessKey/secretKey) to access the default account, you can create buckets in the default account through any interface. - No object access via file interfaces
- Currently, you cannot access objects through file interfaces (NFS, POSIX, CSI). Use
mc
commands to access files in HPE Ezmeral Data Fabric File Store. Themc
commands provide an S3 interface to File Store. - You cannot use mapr. as a prefix when naming buckets
- By design, you cannot name buckets with
mapr.
as the prefix, for examplemapr.bucket1
. - Jobs cannot reclaim space when delete markers exist on versioned objects
- In Object Store, versioning and delete markers work the same as they do in S3 – If
you remove a directory or object from a versioned bucket using the
mc rm
command, a delete marker is placed against that directory or object while all previous versions of the directory or object are retained. These retained versions occupy space that jobs cannot reclaim when they run against the versioned bucket. If you want jobs to reclaim space, you must use the--versions
option to remove all versions of a directory or object when you run themc rm
command, for example:/opt/mapr/bin/mc rm --recursive --force --versions <alias>/<versionedbucket>/<directory>
- S3 Select limitations
-
- The maximum length of a record in the input or result is 1 MB.
- Amazon S3 Select can only emit nested data using the JSON output format.
- S3 select returns a stream of encoded bytes. As a workaround, loop over the returned stream and decode the output records. Example: ['Payload'].decode('utf-8')
- S3 Select only works on objects stored in CSV, JSON, or Apache Parquet format.
- Compressed formats are available for CSV and JSON file formats only.
- Access policy limitation in Object Store
-
- Currently, Object Store only supports the
IpAddress
condition. You can add theCondition
block to a policy, only specifyingIpAddress
within it:"Condition": { "IpAddress": { "aws:SourceIp": [ "192.48.100.222" ] },
- Currently, Object Store only supports the
- Cannot audit data access operations
- Currently, Object Store does not support the auditing of data access operations, as described in Auditing Data Access Operations.