I’m in a hurry with that so please appologize the general topic, usually I only post stuff with a actual issue and added informations.
Anyways, right now I need to decide until tomorrow morning if rockstor is the right choice.
Some things convinced me to believe this, some questions are left.
So to make things short I just write some stuff and you guys might add your thoughts.
This is for a small agency with mac only environment, afp or smb wouldn’t matter, stability is the most important part. Spotlight would be a huge benefit but isn’t time critical.
daily backup on an external harddrive is critical, we plan to have two in a daily cycle.
So far it was raid10 and I’d stick to this. Rockstor would run from an usb stick - of course I’d buy a subscription for $15 yearly to get automatic updates.
In the long run a automatic replication would be interesting, maybe even to public cloud.
Anyways, before this gets too complicated, if somebody could say if some of the above stuff isn’t possible to do with rockstor, I’d need to setup again a plain debian with netatalk.
Thx in advance,
edit: meanwhile I did give it a try in a virtual machine. besides there is some unexpected output on the console (blk_update_request: I/O error, dev fd0, sector 0), but I ignored this assuming this is due to vmware.
so far it works really nice I have to admit. I am wondering though if using the centOS underneath just as a regular centOS works fine or if this produces issues. f.e. yum update and upgrade and such things.
ok now i experienced a really scary thing.
I just shutdown the VM and deleted one of four “pool drives” which of course just had been some virtual disks. now my afp shares are just gone, no hint that something happened, no alert or something, no email notification … just silent fail of a complete afp setup and the pool is still reported as before, now just with three disks in stead of four.
now a polite “wtf??” about this. What is happening here?
Am I totally wrong with testing rockstor in virtual environment?
For your question on the base ya it is a full CentOS core install so you can install do whatever you want.
Stability: CentOS and even the elrepo kernel are pretty stable but both BTRFS and Rockstor are both under fairly heavy development so I would plan updates around that and apply manually for now as well as getting the stable branch subscription. Plus if you do go the USB route make yourself a master img then a bad update is a reimg to fix.
Cloud backup has been discussed in the Rock on requests thread and replication to another rockstor box using btrfs send is built in.
Spotlight: pretty sure this is just the mac search so you shouldn’t have any trouble mounting the drive and then searching it works for windows.
Your /fd0 error is most likely a flobby drive error so for whatever reason you vm thinks there is or was a floppy drive no biggie.
No that is expected behavior (though the reporting could be better). BTRFS won’t automatically mount in a degraded state (disk missing) Thus since the pool backing that share isn’t mounted it can’t make the share available.
If you use the CLI you will be able to mount the array with a mount -o recovery,degraded /dev/sdX /mnt2/pool name
so am I understanding this correct, with btrfs I don’t get a notification if a drive fails?
also it can’t be intended that the afp shares are just gone??
Well there would have been an error in the startup for sure (have to remove plymouth though to see it scroll by) and in the logs etc.
Did you setup an email for notifications in rockstor? I haven’t personally used it myself though since mine is just a home testing/storage box so I’ve been playing kinda fast and loose there so I don’t know if it does send the notification or not.
And for the shares being gone they don’t currently exist to the system since the disks aren’t mounted so there is nothing to share. They will reappear when you mount the array.
Yeah email are setup. Testmail worked.
What I did:
- setup raid10 with 4 vhhd
- setup afp and shares
- shutdown and remove a disk
I can follow your explanation that the shares are kind of hidden, this leads though to questions:
- what will happen if a drive fails on runtime?
- is a raid10 with that not really a raid anymore?
I mean the concept that a drive can fail without a problem is broken by this…
-RAID 10 4 disks
-Remove disk (a single one) while running
-Array continues to function in degraded state, should trigger notification.
It was a design choice by the developers to prevent degraded arrays from being unknowingly mounted. You must explicitly chose to mount it knowing it is in a degraded state.
Degraded in this case meaning the redundancy of the array has been compromised in some way.
Ok I see, this makes sense now
Ok cool this now and your responses now convinced me to give it a try and I’m going to install rockstor this morning.
Only thing to consider now would be if I use usb thumb drive or small ssd, maybe even raid1 of ssds for the OS.
Thx for your responses again, this really is promissing to see an active community backgrounding this project.
Awesome glad you’ve got it now. Let us know how it goes.
It won’t matter too much which one you chose (though I’ll plug their store here with preinstalled USB and mSATA) OS speed won’t mess with the storage speed but it will make administration nicer if you have a faster OS.
Either way I would make yourself a backup master IMG of the drive so that you can easily restore the OS if say you have a update that disagrees with your system or someone changes a settings and goofs it up.
I would also recommend doing some reading on BTRFS so you don’t get any more surprises like it not mounting when you thought it should. The BTRFS and [Arch Wiki] (https://wiki.archlinux.org/index.php/Btrfs) are both good places to start.