Preserving file ACLs, attributes, and timestamps with Azure Data Box
Azure Data Box lets you preserve access control lists (ACLs), timestamps, and file attributes when sending data to Azure. This article describes the metadata that you can transfer when copying data to Data Box via Server Message Block (SMB) to upload it to Azure Files.
Transferred metadata
ACLs, timestamps, and file attributes are the metadata that is transferred when the data from Data Box is uploaded to Azure Files. In this article, ACLs, timestamps, and file attributes are referred to collectively as metadata.
The metadata can be copied with Windows and Linux data copy tools. Metadata isn't preserved when transferring data to blob storage. Metadata is also not transferred when copying data over NFS.
The subsequent sections of the article discuss in detail as to how the timestamps, file attributes, and ACLs are transferred when the data from Data Box is uploaded to Azure Files.
Timestamps
The following timestamps are transferred:
- CreationTime
- LastWriteTime
The following timestamp isn't transferred:
- LastAccessTime
File attributes
File attributes on both files and directories are transferred unless otherwise noted.
The following file attributes are transferred:
- FILE_ATTRIBUTE_READONLY (file only)
- FILE_ATTRIBUTE_HIDDEN
- FILE_ATTRIBUTE_SYSTEM
- FILE_ATTRIBUTE_DIRECTORY (directory only)
- FILE_ATTRIBUTE_ARCHIVE
- FILE_ATTRIBUTE_TEMPORARY (file only)
- FILE_ATTRIBUTE_NO_SCRUB_DATA
The following file attributes aren't transferred:
- FILE_ATTRIBUTE_OFFLINE
- FILE_ATTRIBUTE_NOT_CONTENT_INDEXED
Read-only attributes on directories aren't transferred.
Alternate data streams and extended attributes
Alternate data streams and extended attributes are not supported in Azure Files, page blob, or block blob storage, so they are not transferred when copying data.
ACLs
Depending on the transfer method used and whether you're using a Windows or Linux client, some or all discretionary and default access control lists (ACLs) on files and folders may be transferred during the data copy to Azure Files.
Transfer of ACLs is enabled by default. You might want to disable this setting in the local web UI on your Data Box. For more information, see Use the local web UI to administer your Data Box and Data Box Heavy.
Note
Files with ACLs containing conditional access control entry (ACE) strings are not copied. This is a known issue. To work around this, copy these files to the Azure Files share manually by mounting the share and then using a copy tool that supports copying ACLs.
ACLs transfer over SMB
During an SMB file transfer, the following ACLs are transferred:
- Discretionary ACLs (DACLs) and system ACLs (SACLs) for directories and files that you copy to your Data Box.
- If you use a Linux client, only Windows NT ACLs are transferred.
ACLs transfer over Data Copy Service
During a data copy service file transfer, the following ACLs are transferred:
- Discretionary ACLs (DACLs) and system ACLs (SACLs) for directories and files that you copy to your Data Box.
To copy SACLs from your files, you must provide credentials for a user with SeBackupPrivilege. Users in the Administrators or Backup Operators group will have this privilege by default
If you do not have SeBackupPrivilege:
- You will not be able to copy SACLs for Azure Files copy service jobs.
- You may experience access issues and receive this error in the error log: Could not read SACLs from share due to insufficient privileges.
For more information, learn more about SeBackupPrivilege.
ACLs transfer over NFS
ACLs (and metadata attributes) aren't transferred when you copy data over NFS.
Default ACLs transfer
Even if your data copy tool doesn't copy ACLs, the default ACLs on directories and files are transferred to Azure Files when you use a Windows client. The default ACLs aren't transferred when you use a Linux client.
The following default ACLs are transferred:
Account permissions:
- Built-in Administrator account
- SYSTEM account
- SMB share user account used to mount and copy data in the Data Box
Security descriptors with these properties: DACL, Owner, Group, SACL
Copying data and metadata
To transfer the ACLs, timestamps, and attributes for your data, use the following procedures to copy data into the Data Box.
Windows data copy tool
To copy data to your Data Box via SMB, use an SMB-compatible file copy tool such as robocopy
. The following sample command copies all files and directories, transferring metadata along with the data.
When using the /copyall
or /dcopy:DAT
option, make sure the required Backup Operator privileges aren't disabled. For more information, see Use the local web UI to administer your Data Box.
robocopy <Source> <Target> * /copyall /e /dcopy:DAT /B /r:3 /w:60 /is /nfl /ndl /np /MT:32 or 64 /fft /log+:<LogFile>
where
Option | Description |
---|---|
/copyall |
Copies all attributes. |
/e |
Copies subdirectories, including empty directories. |
/dcopy:DAT |
Copies data, attributes, and timestamps. Note: The /dcopy:DAT option must be used to transfer CreationTime on directories. |
/B |
Copies files in Backup mode. |
/r:3 |
Specifies 3 retries on failed copies. |
/w:60 |
Specifies a wait time of 60 seconds between retries. |
/is |
Includes the same files. |
/nfl |
Does not log file names. |
/ndl |
Does not log directory names. |
/np |
Does not display progress of the copying operation. |
/MT:32 or 64 |
Uses multithreading, with 32 or 64 threads. |
/fft |
Reduces time stamp granularity for any file system. |
/log+:<LogFile> |
Appends the output to the existing log file. |
For more information on these robocopy
parameters, see Tutorial: Copy data to Azure Data Box via SMB
Note
If you use /copyall
to copy your data, the source ACLs on directories and files are transferred to Azure Files. If you only had read-access on your source data and could not modify the source data, you'll have read-access only on the data in the Data Box. Use /copyall
only if you intend to copy all the ACLs on the directories and files along with the data.
Use robocopy to list, copy, modify files on Data Box
Here are some of the common scenarios you use when copying data using robocopy
.
Copy only data to Data Box, no ACLs on directories and files
Use the
/dcopy:DAT
option to only copy data, attributes, timestamps. ACLs on directories and files aren't copied.Copy data and ACLs on directories and files to Data Box
Use
/copyall
to copy all the source data including all the ACLs on directories and files.List the filesystem on Data Box using robocopy
Use this command to list directory contents:
robocopy <source-dir> NULL /l /s /xx /njh /njs /fp /B
Note that the File Explorer doesn't allow you to list these files.
Copy or delete folders and files on Data Box
Use this command to copy a single file:
robocopy <source-dir> <destination-dir> <file-name> /B
Use this command to delete a single file:
robocopy <source-dir> <destination-dir> <file-name> /purge /B
In the above command, the
<source-dir>
should not have the file:<file-name>
. Then, the above command syncs the destination with the source, resulting in the removal of the file from the destination.Note that the File Explorer may not allow you to perform the above operations.
For more information, see Using robocopy commands.
Linux data copy tools
Transferring metadata in Linux is a two-step process. First, you copy the source data using a tool such as rsync
, which doesn't copy metadata. After you copy the data, you can copy the metadata using a tool such as smbcacls
or cifsacl
.
The following sample commands do the first step, copying the data using rsync
.
cp -aR /etc /opt/
rsync -avP /etc /opt (-a copies a directory)