Shared Drives on the same blackplane

[JJ Hardwarez]

I have a Dell SC 4020 i converted in to a truenas server. The SC 4020 comes with LSI 2308 SAS RAID controllers. The SC 4020 has two modules (PC's ) that share the same black plane and drives so when module fails the other is alreay up and nothing lost.

If i run 1 module or the other module, both units by themselves come up fine.. storage pool up and access to the files. If i run both of them at the same time, I dont get any error messages persay in the events dmesg.. but 1 controller has access to the pool, and the other dosent. I also get amber lights on the drives that should be "shared" that they are offline.. but the other module has access to them and are functioning correctly.

below is some photos of what I have going on.. I also changed the NIC cards from 2x 10g sfp to a 40g QSFP card.

thing flys !

Drives showing error (amber) but totally accessible from the 1st booted up module (PC)


Shows Store-A up and running


Drive List





This is Module 2 ( PC ) Store-A no mounted and showing offline



The Drive list also disks are absent..

I did flash the newer 2008ET / IT firmware into both controllers.

Is this an option on the controllers or am i missing something ?

 

[HoneyBadger]

Hello JJ,

Your ZFS pool is active on "Module 1" and as such isn't mountable on "Module 2."

Shared-storage HA requires significant platform customization to make it reliable, and is only available on iXsystems-supplied TrueNAS appliances (like the M-series) with TrueNAS Enterprise - it's not available in TrueNAS CORE.

 

[ChrisRJ]

Shared-storage HA requires significant platform customization to make it reliable, [..]

In other words: There is a high risk of actually reducing availability a lot, if you don't know exactly what you are doing. After all, HA setups add a considerable amount of complexity. Complexity as such adds risk and therefore statistically reduces availability. So the measures that cause this additional complexity, need to be pretty comprehensive for the overall solution to be better than something simpler.

My experience in the enterprise space (although more with transaction systems and not storage) has consistently been that people try to avoid clustering and complex HA setups, if that is possible. A horizontal scaling of web servers with a load balancer is straightforward (at least until we have sticky sessions), but something like Oracle RAC is a different beast. But I digress ...