I'm in the process to migrate 2x Unraid servers with each having 24 drives to newer hardware (2x2 parity, 2x19 data, 2x2 Cache, 2x1 Unassigned Devices). Many services of the second server belong to the first one. Here I use SMB mounts (Unassigned Devices) to include drives from the second server in the first server.

 

My first thought was to keep both seperated. My second thought was: I don't need these duplicate parities and caches if I would put everything in one server. From 48 drives I would save 3 drives (2x Cache, 1x Unassigned Devices). This would give at least two additional scenarios:

 

1.) There are servers with 36, 48, 60 - even 90 - drives nowadays.

 

2.) Expanded storage (like Supermicro, just a powerboard in the storage and a SAS link between the server and the additional storage).

 

There are drawbacks:

 

1.) I would like to keep the parity for max. 22 data drives. This is not possible with the current Unraid OS.

 

2.) Unraid does not allow that many data drives.

 

What's Unraids perspective? What would you guys do? How would you include services of the second server into the first one?

 

Many thanks in advance.

 

What you have not defined for us is 'What data is going to be on two servers as opposed the data that is on one server?".    If one server is to a backup to the other server (even partially), having a single server to where one part is a 'backup' for another part is not a true "Backup".  There are a lot of points of failure that are not related to the data disks that can cause data loss!  

 

Since you already have (or are planning) on dual parity, you might want to read through this thread which contains a discussion of other risks to data.  

 

          https://forums.unraid.net/topic/50504-dual-or-single-parity-its-your-choice/

 

 

Remember the individual servers are in data farms that are part of a very large data array in which parts of it are even scattered over a large geographical area.  Just because they have 90 drives (or whatever number) does not necessary mean that it a desirable choice for you.  Those 90 drives are very small percentage of the total numbers of drives involved in a much bigger configuration! 

Thanks for your questions.

 

I already use dual parity on both servers. The data on both servers is, with the exception of one single disk, not that important. I'm very paranoid if it comes to backup of important data. There are several backups of that single disk in various safes in various places.

 

Back to the not important data. This data is of identical type and spread on all remaining disks on both servers. I would like to access this data from one of both servers only. To do that I do use 54 SMB mount points (Unassigned Devices) to connect server2 to server1. The service (docker) that has access to all servers is running on server1. It uses the SMB mount points to server2 to gain transparent access to both servers.

 

To have all these disks in one server case, or one server and one expander case, would be easier.

 

Having dual parity twice for chunks of disks (say 24) would be ok. But having one server only to handle all 48 disks is the most important point.

 

What would you do? Are SMB mount points thru Unassigned Devices the preferred way to connect a second server? Is there a better way? Or should I hold my breath for upcoming unRAID enhancements? The mouth-watering hardware, I can see on the Supermicro site, needs more features than the 30 data disks that unRAID can handle.

 

I'm currently collecting all information for new hardware. Any ideas are highly appreciated.

 

45 minutes ago, hawihoney said:

 

I'm currently collecting all information for new hardware. Any ideas are highly appreciated.

 

You might want to take a look at this:

 

    https://www.backblaze.com/blog/open-source-data-storage-server/

 

By the way, I found this by googling "backblaze server design" and this was one of many hits...