Here is some follow up on this project. Before doing the in place upgrade I had the system configured as a storage target for XCP-NG with both NFS and iSCSI. I ran series of FIO benchmarks using Phoronix, did the in place upgrade as shown in the video, then ran the same benchmarks again. Here are the results:
Now I have changed nothing on the system at all and have made no attempts yet to figure out why the TrueNAS Scale system was so much slower on for both NFS and iSCSI. But I have had some weird pausing issues that keep coming up on my other system using TrueNAS Scale using SMB. Overall I have not had any crashes that I can point to or things not work, just sub-par performance. I look at all of this as part of the beta testing and I am open to any ideas people might have to get things operating faster.
Those results are very impressive. Do you think it might have something to do with the in-place upgrade – although honestly the only thing being upgraded is really the disks that TrueNAS runs on – the data zfs disks aren’t really going to change.
Perhaps its the beta release.
I’m curious if you posted these numbers over on the TrueNAS forums. You honestly have the first head to head numbers that I’ve seen comparing the two setups on the same machine.
I used this guide to migrate my Truenas Core installation to scale (22.02-RC.1-1). The migration appears to have gone ok in that all of the settings, etc, are maintained and my pools are still present. However, I have run into problems with SMB shares. There is clearly an ACL issue. I have read on the Truenas forums this may have to do with Linux not supporting NFSV4 ACLs. However, changing ACLs to POSIX doesn’t help. There are some posts again on the Truenas forums that suggest either the GUI method of changing ACLs is broken or perhaps can’t even be done with a datastore created in Core and migrated to Scale. Anyway, I don’t think Tom mentioned problems with SMB shares in his video and wanted to bring this up.