N+1 DS-System


Creation Date: August 11, 2006
Revision Date: October 04, 2012
Product: DS-System

Summary

This section contains details about configuring an N+1 DS-System. It is reproduced from the corresponding section in the DS-System Installation Guides (Windows / Linux).

1. The N+1 formation of the DS-System (N+1 DS-System) is designed to give better scalability, and to increase availability for backups. It is designed so that the N+1 DS-System can survive failures of some of its nodes without interrupting the backup service.

2. Configuring an N+1 DS-System means that several DS-Systems will work together to provide backup and restore services to the same DS-Clients. Any of those DS-Systems is able to provide any DS-Client with the same service (backup, restore, delete, synchronization, admin, etc.).

3. To ensure that each individual DS-System's activities are synchronized (i.e. no two activities will conflict), the N+1 DS-System will select one DS-System to act as a synchronization point between all the DS-Systems (called the "DS-Director"). Any DS-System that is not the DS-Director is called a "Node" (or Leaf).

4. To make sure that only one DS-Director is active at any time, a DS-Director will only be created if the DS-Director is connected to "n/2" Nodes. This means an N+1 formation can exist as long as at least "n/2+1" DS-System instances in the N+1 are running and can connect to each other.

This gives an upper bound on how many crashes an N+1 configuration can survive:

5. When "n/2+1" DS-Systems are active, the N+1 configuration will form once the DS-Systems have selected a DS-Director. Once the DS-Director has been selected, the N+1 DS-System starts and all the DS-Systems will accept incoming DS-Client connections (including the DS-Director) and the DS-Director will accept incoming DS-Operator connections.

6. You must synchronize the time on each DS-System that makes up the N+1. You can use UTC (Coordinated Universal Time) via the NTP (Network Time Protocol) server, or any other third party utility that can keep the times synchronized on all DS-Systems in the N+1.

Hardware and Software Setup (N+1 DS-System)

1. All DS-Systems that form the N+1 must have access to the same DS-Client data (in order for each DS-System node to be able to provide the same services to any of the DS-Clients). This means that the DS-System nodes must have access to the same backup root.

2. All DS-Systems that form the N+1 must have access to the same database (containing the DS-Client, Library and other data). This means that a central database must be configured for the DS-Systems to access. It is recommended that this database be clustered.

3. Since each of the DS-System nodes runs on its own machine, each node will have its own configuration file specifying database connectivity, identification inside the N+1, and registration information. In addition, there will be a configuration file for the entire N+1.

4. Install and configure the database server that the N+1 DS-System will use for its dssystem database: N+1 DS-Systems use the same type of database as the standalone DS-System. For the supported database and OS combinations, refer to the file "Installation and Backup&Restore Support.pdf".

Configuring a Microsoft SQL Server database (Windows DS-System):

Configuring a PostgreSQL database (Linux DS-System)

5. Configure the shared storage.

Windows Platform:

The following example uses a UNC path (but the actual implementation is up to you):

Linux Platform:

The following example uses NFS (but the actual implementation is up to you):

6. Install the DS-System on each of the destination node machines:

7. Configure the DS-Systems:

8. Configure the N+1 configuration file ("<backup_root>/cluster/config").

9. Start the DS-Systems.

Connectivity

1. The DS-Operator will only be able to connect to the DS-Director (once the N+1 is active).

2. A DS-Client can connect to any of the DS-Systems (including the DS-Director) for any activity. The DS-Client is able to accept multiple DS-System IP addresses to connect to (as it is the case for the N+1) and will perform load-balancing of connections by itself. Another option is to use a hardware load-balancer and a single IP address.

Frequently Asked Questions (FAQ) about N+1 DS-Systems

1.
Question
How is a Node crash handled by the N+1?
Answer
  • In case a Node crashes (for example because of hardware failure), all activities on that Node will be interrupted. However, if there are still "n/2+1" DS-Systems active, the rest of the N+1 still functions and the rest of the activities running on other DS-Systems continue to run. In addition, in case the activities that were interrupted were scheduled, they will retry in 5 minutes to connect to the DS-System and will succeed since they will connect to another DS-System in the N+1 formation.
2.
Question
How is a DS-Director crash handled by the N+1?
Answer
  • In case the DS-Director fails, the DS-Systems will have no synchronization point and to avoid any problems, they will stop any activities and will try to find another DS-Director. Once a new DS-Director is elected, the N+1 will start again.
  • Because of the importance of the DS-Director, any Node that loses connection with the DS-Director will stop activities and move to stand-by mode.
3.
Question
How many DS-Systems can I have in an N+1 configuration? Is there a minimum or a maximum?
Answer
  • A minimum N+1 would probably be 3 DS-Systems, since it can survive the crash of 1 DS-System. A N+1 of 2 DS-Systems is theoretically possible, however it does not make any sense since in case 1 system crashes, n/2+1 DS-Systems are not running and the N+1 will move to stand-by mode.
  • There is no actual maximum. It can be anywhere from 3 to 50 DS-Systems or more. However it must be kept in mind that each Node must have a DS-Director connection (see the star-shaped connectivity in the "N+1 Status" dialog). This means that the higher the number of clustered DS-Systems the more load the DS-Director is required to handle and more communication will go to/from the DS-Director.
4.
Question
How do I add a new DS-System to the N+1? Can I simply edit the N+1 configuration file?
Answer
  • No. A new N+1 license is required: from a DS-License Server.
  • On the DS-License Server, you only need to add the new node's IP address to the existing N+1 DS-System.
  • You can add the node to the N+1 configuration (via the DS-Operator GUI). See "Add an N+1 Node".
  • You must manually edit the new node's local "dssys.cfg" file to add the "Cluster ID : <number>" line, then you can start the node and it will automatically join the N+1.
5.
Question
How do I delete a node from an N+1 DS-System? (e.g. changing from a 4-node N+1 to a 3-node N+1)
Answer
  • Stop the DS-System service / daemon on the node you want to remove.
  • Update the license to the lower number of nodes. This is done from the DS-License Server.
  • Remove the node from the N+1 Cluster Config file using DS-Operator > N+1 Menu > Status > List Tab: Delete Node. See "Delete an N+1 Node".
6.
Question
Do I have to stop the N+1 each time I replace a crashed DS-System?
Answer
  • No. If the N+1 configuration file is not changed, a DS-System can simply be installed and started (make sure no duplicate ClusterIDs are in the DS-System configuration file). The DS-System will join the existing N+1 by itself.
7.
Question
What license is required to run N+1?
Answer
  • On the DS-License Server, you must specify that the license is for an N+1 DS-System configuration.
8.
Question
What happens if the date or time is not synchronized on an N+1 DS-System?
Answer
  • You will not be able to track history in the Activity Log or Event Log.

See Also



The information provided in this document is provided "AS IS", without warranty of any kind. ASIGRA Inc. (ASIGRA) disclaims all warranties, either express or implied. In no event shall ASIGRA or its business partners be liable for any damages whatsoever, including direct, indirect, incidental, consequential, loss of business profits or special damages, even if ASIGRA or its business partners have been advised of the possibility of such damages. © Asigra Inc. All Rights Reserved. Confidential.


PREVNEXT