How do NDMP session limitations work in ONTAP?

Applies to

• ONTAP 9
• NDMP

Answer

This article aims to answer the following NDMP session scalability questions for storage controllers running ONTAP

1. What are the different types of NDMP sessions?
2. What are the platform's NDMP session limits?
3. How do these limitations work with cluster-wide SVMs (vServers) and LIFs?
4. How can I monitor my NDMP sessions?

What are the different types of NDMP sessions?

In order to understand NDMP session limitations, it is important to know how NDMP sessions are classified. NDMP sessions can be broken up into two categories:

1. Dump and Restore sessions: These are NDMP sessions directly responsible for backing up data to and from disk or tape. They interface with the NDMP data or tape service.
2. Control or 'common' sessions: These are bi-directional NDMP control connections responsible for NDMP control commands or messages between the NDMP client (typically a backup application) and the NDMP server (typically a NetApp Storage Controller).

Dump and Restore sessions are platform-dependent in Data ONTAP 8.2 and later releases and can be seen in the table below. Control sessions are hard-coded to 4 regardless of platform, but can be increased manually using the following advanced command. This is typically used if a backup application opens several control sessions for one dump or restore session.
::> set advanced
::*> system services ndmp modify -node <node_name> -common-sessions <integer>

Note: The above command applies regardless of the NDMP being node-scoped or vServer-scoped.

For more information, see the system services ndmp modify manual pages.
 

What are the platform's NDMP session limits?

NDMP session limits ONTAP are based on platform memory and are per-node values. If you are unsure in which category your controller belongs, refer to Hardware Universe.

Memory	Control/common session limit	Dump/restore session limit	Total session limit
WAFL low memory platform or <16GB	4	4	8
WAFL mid memory platform or >=16GB and <24GB	4	16	20
WAFL high memory platform or >=24GB	4	32	36

How do these limitations work with cluster-wide SVMs (vServers) and LIFs?

NDMP session limits are per-node limits. Therefore, each node has its own NDMP session 'pool'. Which pool is used depends on which node is hosting the LIF being used for the NDMP connection. Due to this, it is important to keep in mind what LIFs you are using for NDMP connections and on which nodes these LIFS are hosted.

For example, consider a scenario of a 2-node cluster of FAS3250s running clustered Data ONTAP 8.2. This platform has a total NDMP session limit of 20. The cluster admin creates SVM1 with data_LIF1 and SVM2 with data_LIF2. If data_LIF1 and data_LIF2 are both hosted on node1 in the cluster, all NDMP sessions created through these LIFs will share the same session pool of 20 total sessions.

Note: NDMP session limits are independent of NDMP scope.

 

How can I monitor my NDMP sessions?

Follow these steps to find the total number of NDMP sessions running in clustered Data ONTAP 8.2 and later.

1. Run system services ndmp node-scope-mode status to determine in what scope NDMP is currently operating on the cluster. If enabled, the cluster is operating in node-scoped mode. If disabled, the cluster is operating in vserver-scoped mode.
   ::> system services ndmp node-scope-mode status
NDMP node-scope-mode is disabled.


2. If operating in node-scoped mode, run system services ndmp status to see the open NDMP sessions on the cluster. If operating in vserver-scoped mode, run vserver services ndmp status.
   cluster::> vserver services ndmp status
                     Session
Vserver                Id
------------------ ------------
cluster            1000:60293
cluster            1001:67917
2 entries were displayed.


3. Use system services ndmp probe if in node-scoped mode or vserver services ndmp probe along with the NDMP session ID from the previous command to find more data about an NDMP session. If there is an active Data State or Mover State, this is an NDMP dump or restore session. This command will also display on which node the operation is running.
   cluster::> vserver services ndmp probe 1001:67917

Vserver: cluster
Session Identifier: 1001:67917
NDMP Version: 4
Session Authorized: true
Data State: ACTIVE           <--ACTIVE Data State indicating dump or restore session
Data Operation: RESTORE      <--Data Operation is a RESTORE
Data Server Halt Reason: NA
Data Server Connect Type: TCP
Data Server Connect Address: 169.254.0.1
Data Server Connect Port: 52797
Data Bytes Processed: 350859264
Mover State: IDLE
Mover Mode: NOACTION
Mover Pause Reason: NA
Mover Halt Reason: NA
Mover Record Size: 0
Mover Record Number: 0
Mover Bytes Moved: 0
Mover Seek Position: 0
Mover Bytes Left to Read: 0
Mover Window Offset: 0
Mover Window Length: 0
Mover Position: 0
Mover SetRecordSize Flag: false
Mover SetWindow Flag: false
Mover Connect Type: LOCAL
Mover Connect Address:
Mover Connect Port: 0
Effective Host: PRIMARY
NDMP Client Address: 169.254.0.1
NDMP Client Port: 29773
SCSI Device ID: None
SCSI Host Adapter: -1
SCSI Target ID: -1
SCSI LUN ID: -1
Tape Device: Not open
Tape Mode: READ
Node: cluster-02             <--NDMP session is established on node "cluster-02"
Is Secure Control Connection: false
Data Backup Mode: DUMP
Data Path: /dataSVM/restoreVol
NDMP Source Address: 169.254.0.1

Additional Information

N/A