TrueNAS Scale (HA) How to?

[webix]

Hello Folks.

Since TrueNAS announced that TrueNAS Scale can now do High-Availability (https://www.truenas.com/blog/truenas-scale-clustering/), I want to know if someone effectively did it.
I have 2 nodes that I want to setup has federated configuration (or mirrored cluster, wich one is easier to configure):

I have searched a lot here on the forums and on internet and didn't found anyone that effectively did it. And I found it strange that TrueNAS announces that TrueNAS Scale can be configured like any of these types above without giving the info on how to do it.

The config should go like this:

Any help on setting this up is appreciated.

 

[danb35]

I don't have experience about actually doing it, but I understand it would need to be done through TrueCommand. You might find more information in connection with that.

 

[Ericloewe]

HA is only on suitable iX hardware:

Shared storage HA requires significant platform customization to make it reliable and is only available as TrueNAS Appliances. TrueNAS SCALE can be used on the TrueNAS M-Series platform and deliver HA for both storage services and for its VMs and container workloads. This update is supported since SCALE 22.02.2.

The federated stuff is probably done at a higher level, not sure if TrueCommand can help with that.

For the Mirrored Cluster, you actually need a minimum of three servers, if you want any semblance of reliability and data protection. Gluster sucks like that. You would configure it with TrueCommand - or manually like you would a "standard" system.

 

[webix]

Hello @danb35,
Thank you for your reply. After some research on Discord I get to the same info.

Hello @Ericloewe,
It's not what the announcement says! iX hardware is for TrueNAS Enterprise only. The announcement talks specifically about TrueNAS Scale.

 

[webix]

@Ericloewe also, I am not talking about shared storage but mirrored storage.

 

[Ericloewe]

Hello @Ericloewe,
It's not what the announcement says! iX hardware is for TrueNAS Enterprise only. The announcement talks specifically about TrueNAS Scale.

I quoted the announcement directly.

@Ericloewe also, I am not talking about shared storage but mirrored storage.

See above:

For the Mirrored Cluster, you actually need a minimum of three servers, if you want any semblance of reliability and data protection. Gluster sucks like that. You would configure it with TrueCommand - or manually like you would a "standard" system.

 

[morganL]

Hello @danb35,
Thank you for your reply. After some research on Discord I get to the same info.

Hello @Ericloewe,
It's not what the announcement says! iX hardware is for TrueNAS Enterprise only. The announcement talks specifically about TrueNAS Scale.

The HA functionality is tied to special clustering software and specifically engineered hardware... that includes PCIe Non-Transparent-Bridging and dual port SAS/NVME media. This ensures that we can deliver very good latency and efficiently us the storage media. These configs support all protocols.

HA Is only currently supported for our Enterprise appliances - TrueNAS M-Series and X-Series. Its difficult to validate and troubleshoot, so we only deploy on iX hardware with lots of engineering time and many lab deployments.

These iX appliances can run TrueNAS Enterprise software or can now run TrueNAS SCALE. TrueNAS SCALE can effectively have the Enterprise/HA license included when using that hardware. The TrueNAS SCALE HA includes VM and Apps as well.

For Community users we recommend to start with a 3 or 4 node cluster.
3 Node clusters support Gluster/SMB with ability to sustain a single node failure.
4 Node clusters can support an S3/Minio cluster as well.

 

[HoneyBadger]

The config should go like this:
View attachment 59343

Any help on setting this up is appreciated.

What you're looking for here, based on the diagram, is either a highly-available iSCSI volume or NFS export, with rapid failover between nodes - and I can't recall if that's in SCALE at this point. It's available in TrueNAS Enterprise as stated by @morganL - and the challenges of making this work across a mirrored cluster include an increase to the latency, as now you have to wait for synchronous writes to be acknowledged by the mirrored cluster node.

There used to be a roadmap that showed when (in terms of "which target version") certain features were to be made available for testing and release, but I can't find it now - I did see the Jira roadmap, but I was thinking of something different that had the 22.xx releases and the "feature targets" for each.

 

[morganL]

I should add that Mirror cluster storage for SMB has not passed its reliability tests. It can be manually configured, but it is not recommended.

Only 3 node setups have proven reliable when there are major power or network events.