NDMP and DAR

Direct access restore [DAR] is the ability to restore specific files without having to go through the ENTIRE TAPE. DAR provides the EXACT offset at restore time to the backup application. With this information, the backup application is able to JUMP directly to the DESIRED OFFSET and recover the file without having to read the ENTIRE backup IMAGE sequentially.

DAR requires:
1. NDMP version 3 or later.
2. File History [HIST environment variable) to be enabled.

Note: "File History" is the term used for an INDEX of files that has been backed up. An NDMP client [Backup Application] may request FILE HISTORY from the NDMP Data Service at the time of initiating the backup. If the file history is requested, the backup engine sends information like -File name and path, file status information and file positioning information (i.e. the address of the file in the backup data stream).

It is important to note that the "File History" and "DAR" are not the same, and this is b'cos you can choose to restore a file without the 'File Positioning Information', and this will be called Non-DAR restore, in other words, you have an ability to browse the folders/files and then select the specific file(s) to restore, but a large portion of the data from the backup that included the file must be read, could well be the entire tape.

Whereas, in DAR : Only the portion of the TAPE which contains the data to be restored is read.
Hence, they both are two different things.

This further means that the offset map [DAR] file generation can result in a significant impact to performance if a large number of small files are backed up. In such a case, you may want to disable the offset map generation WITHOUT disabling the FILE HISTORY and there by increasing the performance and yet have the ability to browse folders/files.

So, Keep FILE HIST=T, and disable DAR [offset_map]

To do so, on the NetApp console, use the following options command.

7-mode: [Disable offset_map]
7-mode> options ndmpd.offset_map.enable off

ONTAP [cmode] -  [Disable offset_map]
::>vserver services ndmp modify -vserver vservername -offset-map-enable  [true|false]

Directory DAR:
Is the ability to provide a directory name to the backup application such that the backup software EXPANDS the directory contents and recovers each FILE using the DAR process.

To be able to implement Directory DAR, ONTAP needs to record the offset of each file on the backup image. Thus, on restore time the restore application can EXPAND the directory to be recovered, and then LOAD the information about the offsets, and do DAR for all files underneath the specified DIRECTORY.

Enhanced DAR: Is nothing but combination of FILE HISTORY & OFFSET map both enabled .Thus to benefit from the Enhanced DAR feature, HIST & OFFSET must be enabled.

NDMP DUMP methods, walking inode file Vs logical traversal

There are two methods dump uses to determine which file goes into a dump - the "walking inode file" path and the "logical traversal" path. 

Walking inode file: This approach goes through all the inodes in the inode file and decides which file goes into the dump.

Logical traversal :This approach does a logical traversal on the subtree to be dumped. When dumping a qtree tree, the first approach is used. 

However, under the certain conditions, it becomes more expensive to do "walking inode file" than "logical traversal".

For example - When the volume has a lot of used inodes; and The qtree is very small relative to the size of the volume it resides in.

In Cluster data ontap or ONTAP 9, you can determine which method dump is currently set to by entering the following command -

cluster_ontap_9::*> vserver services ndmp show -vserver vserver_name

This switch determines the two methods as shown in figure below.
switch = [-dump-logical-find <text>] (privilege: advanced)

This option mentioned above, specifies whether to follow inode-file walk or tree walk for phase I of the dump. Choosing inode-file walk or tree walk affects the performance of the dump.

This option [-dump-logical-find <text>] can take following values:

If default is specified [which is the default setting], then level 0 and incremental volume as well as qtree dumps will use inode walk.

If always is specified, all the subtree dumps will use tree walk, and all dumps will follow treewalk.

NDMP dump backup and NDMP dump Levels

NDMP dump backup and NDMP Levels on NetApp FAS

Dump is a Snapshot copy-based backup and recovery solution used by NetApp to back up files and directories from a Snapshot copy to a Disk/tape device.

You can back up your file system data, such as directories, files, and their associated security settings by using the dump backup. You can back up an entire volume, an entire qtree, or a subtree that is neither an entire volume nor an entire qtree.

When you perform a dump backup, you can specify the Snapshot copy to be used for a backup. If you do not specify a Snapshot copy for the backup, the dump engine creates a Snapshot copy for the backup, and after the backup operation is completed, the dump engine deletes this Snapshot copy.

NDMP = Mechanism + Protocol
M= dump, tar, cpio
P= TCP/IP + XDR

NetApp use = dump

Ontap 7-mode dump Levels:
With ontap 7-mode : You can perform level-0 Full, incremental[1-9], or differential[1] backups to tape/disk by using the dump engine.

Level 0 or Full Backup:
A Full Backup provides a backup of all the data in the selected path.

Level 1 through 9 or Incremental Backup:
The Incremental Backups base themselves on the most recent lower-level Incremental Backup, and include any data that has changed or is new since the last Full or Incremental Backup. 

Maximum number of consecutive incremental backups permitted after a full backup is 9, as shown below.

After a differential backup, the maximum is 8, as shown below.

dump levels on Clustered Data ONTAP 8.3 onwards:

Clustered Data ONTAP 8.3 onwards supports 32 levels of dump Backups.

Level 0 = is a Full Backup.
Level 1 through Level 31 are Incremental Backups.

Maximum number of consecutive incremental backups permitted after a full backup is 31, as shown below.

Please Note: For data ontap versions prior to 8.3,the maximum number of consecutive incremental backups permitted after a full backup is 9. After a differential backup, the maximum is 8 just like 7-mode.

For IntelliSnap [SnapDiff] NAS backup, there is no such limitation, whatever is the Volume snapshot limit applies, for NetApp systems, a volume can have maximum 255 snapshots.

Could not open an NDMP connection to host. Please verify the NDMP server is running on that host when adding NAS iDA to CommVault v10 & 11.

Following error is seen when adding a NAS iDA NetApp 7-mode client in the CommCell v10.

Customer reported : Telnet to Port 10000 on the NAS Host works.

Telnet to Port 10000 is a good troubleshooting step, however in cases like this, it isn't helpful.

Reason : Telnet to Port '10000' only signifies NDMP server is listening, but we need to find out why is it refusing to speak, and reporting this error - 'Please verify NDMP server is running on that host'.

Cause: NDMP server [FILER] is definitely listening on Port 10000, but it is refusing connection requests, b'cos its NDMP memory pool is Full.

Initial troubleshooting steps:
Ensure NDMP is turned on.
FILER>options ndmpd.enable
ndmpd.enable           on

That indicates, NDMP server is definitely listening, however there is something which is preventing the communication.

First clue, as reported in EvMgrS.log: Connection refused by the NDMP server

This is the first clue, which kind of give us an indication that the NDMP server is unable to accept any more connections.

Second clue, look for stale NDMP sessions on the filer:
FILER>backup status
This will tell you if there is any existing NDMP backups running on the filer. In this case, there were none.
FILER>ndmpd status

In this particular case, we found there were lot of stale NDMP sessions sitting idle and doing nothing, in other words, they were simply holding on to NDMP memory and hence stopping NDMP to accept new connection.

Please note, If there is nothing in the backup status output, then you can safely kill all the stale ndmp sessions that are shown via ndmpd status output.

Solution:
1. Kill all the stale sessions.
FILER>ndmpd killall

2. Turn off and turn On NDMP [This step may be required, but not necessary]
FILER>options ndmpd.enable off
FILER>options ndmpd.enable on

3. Try to add the NDMP NAS iDA client once again, this time it should succeed.


Please Note: For more information on available ndmpd commands, simply type 'ndmpd' on the putty console for 7-mode;
FILER> ndmpd
usage:  ndmpd [on|off|status|probe [<session #>]|kill <session #>|killall|