Question

Doubt about MA architecture

  • 10 March 2023
  • 6 replies
  • 238 views

Badge +3

I will soon have to implement Commvault on a site consisting of the following elements:

- 2 MediaAgent
- A NetApp storage array presenting file per block via iSCSI

The idea is that the MAs work on the same storage and that when one goes down the other continues to provide availability to the infrastructure for both backup and restore operations and reading a lot I think that the best for my case would be to implement GridStor: https://documentation.commvault.com/2022e/expert/10842_gridstor_alternate_data_paths.html

For my knowledge and experience, reading it makes me a somewhat complicated configuration and I don't know if it fits the needs I am looking for.

This is the procedure that is the most difficult for me because I do not understand very well why: https://documentation.commvault.com/2022e/expert/9788_san_attached_libraries_configuration.html

I thought that presenting the LUN to the MA and formatting it and sharing the data path with the other MA would be enough but I see that it is not so. Can someone give me some guidance to make a correct and optimal design?

Now comes the other doubt and it is that in another site I have the following infrastructure:

- 1 MediaAgent
- A NetApp storage cabinet that presents file per block using iSCSI.

What would be the optimal way to present the storage to this MA? Presenting the LUNs and formatting them in NTFS so that it sees them as local disks or following the same procedure to add a SAN storage as in GridStor even though I only have one MA: https://documentation.commvault.com/2022e/expert/9788_san_attached_libraries_configuration.html 

I have tried to look for good block storage practices but I can't find much.

Sorry for the length of the post and thank you very much in advance.
 

Regards


6 replies

Userlevel 4
Badge +10

Hello @eduardo.jimenez 

Can you present the storage as NAS? I see the word NetApp and would not use anything other than CIFS if you can (assuming you are using windows MAs).

Setting up mount paths via LUNs will always result in some sort of limitation as the paths will not be shared across the MAs when one is down. The only way you could do this is to allow mounts to be visible on both nodes at the same time, which may not be possible, or using clustering (not one I would recommend). It is easier to allow backups to continue when an MA is down, but restores may be an issue if the storage is only available on one MA.

Using CIFS is easy, shareable and resilient. You just need to ensure your controllers have enough grunt to cope with the CIFS overhead and data size.

Badge +3

Hi @Graham Swift

First of all thank you very much for taking your time to read and answer and now I will answer your questions and also putting on the table some more :)

For this implementation, a block-only storage device has been purchased, so the NAS part is completely out of the question.

You say that with LUNs everything is more limited but is it possible and if so how would the configuration be? I understand that following the GridStor part, right?

The part of clustering the MAs, is there any CV functionality to make a cluster or should it be a Windows cluster?

Best regards.

Userlevel 7
Badge +19

So I assume you are dealing with a NetApp E-series, right? With the use of iSCSI towards a disk-based storage library you will not be able to use GridStor. This is targeting tape libraries and the mount path sharing via network is something that will still require the source MA to be present to be able to deliver the mount path to the other MA(s).

It is possible to install and configure the MA to run as part of a Windows cluster configuration leveraging WSFC, however it does add-up complexity and most likely will always require a DDB-repair in case the DDB was active during the failover. If HA is one of the requirement than normally I go for a solution that offers a S3-compatible (cloud storage) interface or SMB/CIFS.

Badge +3

So I assume you are dealing with a NetApp E-series, right? With the use of iSCSI towards a disk-based storage library you will not be able to use GridStor. This is targeting tape libraries and the mount path sharing via network is something that will still require the source MA to be present to be able to deliver the mount path to the other MA(s).

It is possible to install and configure the MA to run as part of a Windows cluster configuration leveraging WSFC, however it does add-up complexity and most likely will always require a DDB-repair in case the DDB was active during the failover. If HA is one of the requirement than normally I go for a solution that offers a S3-compatible (cloud storage) interface or SMB/CIFS.

Hi @Onno van den Berg,

Thank you very much for your answer. 

To be more specific the storage device is a Lenovo ThinkSystem DE4000H disk based, yes. We cannot use another device as it is the one purchased for this purpose.

When I was reading about GridStor and the SAN libraries part: https://documentation.commvault.com/v11/expert/9788_san_attached_libraries_configuration.html it is specified that it can be used for libraries on disks as in my case or at least that is what I understand from reading:

 

Sorry to be insistent, it is an advanced configuration for my experience with CV, but I want to understand why you tell me that it is not possible and in the documentation you refer to disk library.

If the configuration with Gridstor is not possible in the end, what would be the best architecture and if it is possible to give a HA to the same one without much complication as you comment with the Windows cluster? And if the MA HA is not possible, what would be the optimal architecture for the two MA in the same site?

Best regards and thank you very much!

Userlevel 7
Badge +19

The link in your post redirects to a page displaying an article showing documentation related to SAN attached libraries, so I assume you meant to link to the following page: https://documentation.commvault.com/2022e/essential/87485_sharing_disk_access_paths_between_multiple_mediaagents.html

Now to be clear: sure there are ways to share the disk libraries through one of the MediaAgents, but if that MediaAgent stops working both MediaAgents will loose access to the library. It will not fulfil HA which is also not possible as you are writing to a block-based device and without a proper clustered file system (like VMFS) that supports concurrent systems to access the block device this unfortunately will not work. So, clustering the MediaAgent is a possibility to add HA. By dividing the available space you could build the setup as such that both MAs actively fulfil a role because otherwise you will have a active/passive setup. 

I hope this information answers your question. 

Badge +3

The link in your post redirects to a page displaying an article showing documentation related to SAN attached libraries, so I assume you meant to link to the following page: https://documentation.commvault.com/2022e/essential/87485_sharing_disk_access_paths_between_multiple_mediaagents.html

Now to be clear: sure there are ways to share the disk libraries through one of the MediaAgents, but if that MediaAgent stops working both MediaAgents will loose access to the library. It will not fulfil HA which is also not possible as you are writing to a block-based device and without a proper clustered file system (like VMFS) that supports concurrent systems to access the block device this unfortunately will not work. So, clustering the MediaAgent is a possibility to add HA. By dividing the available space you could build the setup as such that both MAs actively fulfil a role because otherwise you will have a active/passive setup. 

I hope this information answers your question. 

@Onno van den Berg 

Thanks for the clarification. Having explained everything and clarified, my question is the following: what configuration would be recommended and that performs a correct function with two MA on the same site and both connected to the Lenovo by iSCSI? Could you pass me documentation about it if there is any because I have searched and did not see anything since I always end up in GridStor. Waiting for your answer I go ahead and understand that it will be an active/passive configuration of MA in which if one goes down the other comes into operation, right?

I insist on adapting to this configuration since two new servers have been acquired for the MA role and the storage device and that is what I have to adapt to.

Thank you very much!

Reply