Just a heads up: regardless of the 3.5.5 or 3.7.9 ISO, if you try using a drive larger than 16TB at install time, it defaults to only allowing 16TB. And if you try to manually configure it, the installer will crash with an odd Python error, as others have complained about the CentOS 7 installer. I’ve been attempting to work around it by disabling the auto-provisioning in the kickstart file: but using anything other that DD or RAWRITE to make the flash drive, causes that to become non-bootable due to a missing kickstart file.
I am curious about your hardware setup. How did you provision a drive that large? I wonder if it’s going to be feasible for me to reproduce this problem.
RAID6 array on Dell H700 controller: 4TB x 8 drives. I’ll have to give it a try with CentOS vanilla and report back: I figured the autosizing was maxing out at 16TB because older systems appear to have a 16TB limit for EXT4.
System also has 64GB of RAM: which means unless I use an SSD-card, the installer can’t load the OS onto a flash drive because I can’t disable the default amount of swap, due to the other bug. I originally spec’d this unit for FreeNAS use, but our needs changed, so I’m seeing if Rockstor can handle a reduced role.
Well, it seems CentOS 7.1503 can manually partition just fine: had extracted the ISO to the installer flash drive using Rufus successfully. Can Rockstor be added to a working CentOS setup by any chance?
It can, but it’s not trivial. I want to take the opportunity to improve Rockstor installer. Let me give that a try.
(nods) its duly appreciated!
BTW I meant no offense with my earlier statement about “reduced role” : I was going to be attempting some ZFS replication + maybe have the unit as a hot-spare; now its going to be strictly a target for pre-encrypted data + VM/Xen backups.
None taken, but thanks for the clarification.
Sorry to Necro this thread, but I wanted to know if Michael created the RAID with the controller BIOS or with BTRFS? I am under the assumption that RAID cards should be used with next generation filesystems in IT/HBA/JBOD mode, and arrays/virtual disks should not be created with the card. All disks should be presented as individual physical drives so BTRFS can make the best use of them. I am also informed that most hardware RAID configurations (even if they are a single disk RAID0) will not pass all SMART data through to the OS and drivers along with creating inconsistent performance by interfering with the disk and controller caches.
Please correct me if I am wrong.
Yes, volume management by BTRFS has better self healing and recovery support to say the least. I think some users are still weary of BTRFS and may go with hardware RAID, but I don’t really see clear advantages of that approach in the context of Rockstor. Of course not hardware RAID features are the same, but I am speaking generally.