Amahi Forums

This may be a dumb question - but is there any official guidance on how to deconstruct your AMAHI server in order to do upgrades?

To clarify - the recommendation always seems to be to 'Install From Scratch' - which is fine, but I spent HOURS setting up and configuring disks, letting greyhole sync balancing things out etc - and I really don't want to lose any of that data or create some big mess I'll have to manually untangle.

I don't mind uninstalling or reinstalling apps etc - my main concern is the greyhole configuration. Is there any guidance on how to go about doing a 'fresh install' without borking up the drive config for Greyhole? (Can I just unplug my data drives while installing? If so, when would I reconnect them? Do I have to do it in a certain order? - I have 9 drives I'm using for greyhole and I'm paranoid about losing any of the data as I just migrated everything from my Windows Home Server which is not running anymore.)

You might try the HDA OS Migration Guide.

Although we have not added Fedora 23 to it, the steps will be pretty much the same.

Awesome. I've managed to run the migration script to get the nicely formatted list of everything.

Quick question though:

I originally built this with a HW Raid5 of 4 SSD's for the base system. Gives me about 500GB of more or less wasted space. Then I followed the steps to move /var/hda/files to a 2.5TB drive, which is now basically just a oversized LZ for greyhole.

If I'm reinstalling anyhow, I'd kind of like to break the raid up and reconfigure as one 128GB mirrored pair for system and one striped pair of SSD's for a 512GB super fast LZ for greyhole.

That would free up the 2.5TB drive to add in to greyhole which seems like a much better use of it.

Do you have any recommendations as far as the safest way to go about doing this? (I'm confident I know what I want to do, just not what the safest order to do them in... )

If I'm reinstalling anyhow, I'd kind of like to break the raid up and reconfigure as one 128GB mirrored pair for system and one striped pair of SSD's for a 512GB super fast LZ for greyhole.

Sorry I cannot be much help here as I don't have any experience with RAID. Best I can do is remind you to back up your files somewhere in case there are issues.

I really don't see the point of RAID if you are using Greyhole. Is there that much of a speed increase? In my opinion, the less complicated you make the server setup the better it will be.

I finally got moved and finished my upgrade. There were a few snags in the install that caused me some problems, but they turned out to be easy once I narrowed it down.

1st: Regardless of which f23 install you download and use unetbootin to create a bootable USB key with - all of them failed to install for me and kept dropping into recovery. I immediately thought it was something to do with my hardware and started disconnecting everything and REALLY simplifying my setup to no avail. I finally determined it was an issue with the name (label) of the USB keys I was using - for some reason unetbootin (on mac at least) expands the installer, creates everything just fine but does not change the label of the key to match that expected by the installer. On further probing, the name the installer wanted was too long anyhow so I couldn't rename it to match. Easy enough to resolve, as you can just hit tab on the installer menu and change the string after "LABEL=" to match the name you chose when you formatted your USB key. (Leave the 'quiet' option, just change the name)

If we have an opportunity to modify the distro at all, changing the LABEL value to something shorter would be a VERY good idea. FEDORA_23 is about as long as my mac would let me make it, it shortened anything longer to that anyhow.

Bigfoot: In regards to RAID - I've managed enterprise server farms for almost 20 years... my experience is that all things fail eventually - usually at the least convenient times... So anything I can do in the build to either minimize my exposure or make the recovery easier the better.

I wanted to be able to lose a drive and have the OS itself keep on trucking, so I used a mirrored pair. In this case, I had 4 128GB SSD's in two matched sets that I had from a previous project so I made use of them to add a little protection and performance at no extra cost.

My final setup was this:
Raid1 (Mirror) of 128GB for Fedora Install. (1st pair of SSD's)
Raid0 (stripe) of 256GB for /var/hda/files (2nd Pair of SSD's)
- Wasn't comfortable with only 128GB as a landing zone for greyhole so I striped this set to give enough room to buffer large transfers. Since nothing 'lives' on this set permanently, it doesn't matter that if a disk fails I lose /var/hda/files until I get a chance to plug something else in and configure it. I tested and the system doesn't crash when it's removed - you just can't copy new data to server (which is fine, as greyhole would get out of sync anyhow). I initially mounted this set as /var/hda/drives/LZ to make sure everything looked good then followed the 'move all shares' directions in (https://wiki.amahi.org/index.php/Adding ... o_your_HDA to mount permanently

After that I just mounted all of the other drives I wanted in my storage pool and configured them to mount as:
/var/hda/drives/ghdisk0
/var/hda/drives/ghdisk1 ....

I also made sure to annotate the /etc/fstab to reflect where each drive was physically mounted and the size, model,serial# etc to make sure everything was repeatable if necessary. I'm OCD-ish so I physically labelled each drive and cable as well. (My thought was if all goes well this will live in the basement under the stairs and not physically be touched possibly for years so better to make it as idiot proof as possible to give future self a better chance of fixing whatever goes wrong.)

Anyhow - I'm back up and in business - ran the fsck last night and greyhole re-assembled the data for all of my shares etc. Everything so far looks golden.

Thank you-

Jon

To clarify: Raid is only for OS and LZ. Rest of drives are non-Raid and Greyhole 'manages' them.

Happy to hear you were able to get it all sorted out.

In regards to RAID - I've managed enterprise server farms for almost 20 years... my experience is that all things fail eventually - usually at the least convenient times... So anything I can do in the build to either minimize my exposure or make the recovery easier the better.

Thanks for the clarification. Many don't have your level of experience or drive to implement RAID. If you know it well, why not use it.

For those who do not, I always caution them as we have limited support in the forums for such an advanced configuration.

Closing this thread.