I've got an unassigned disk set up with btrfs that I've been using... poorly.  I have a qcow2 VM image on it, as well as my docker.img.  It has uncorrectable errors now, either due to qcow2 or some instability in my system (fixed with a new motherboard). 

 

My system has been running btrfs since the early days of v6, and I haven't been paying much attention since (thus not seeing qcow2 == bad).

 

I see that btrfs for an SSD cache pool is recommended for VM/docker + redundency (well, I guess it is SSDs that are recommended https://lime-technology.com/exciting-new-developments-with-unraid-6/).  Is btrfs the preferred filesystem, specifically for VMs?  Or something else?

 

Also, for a cache pool, I assume you need to install both disks at once (since it is raid1)?  Or can you add a 2nd drive later?

 

Thanks!

It all depends on who you ask. I currently do not recommend BTRFS at all, and recommend avoiding it.

I'd try it on an array. You can add the second drive at any time. You can only correct certain errors with an array of BTRFS drives, likely the errors you seeing. Even LimeTech didn't know this until I pointed it out a little while ago. I never had much luck with BTRFS on a single cache drive. I get errors in the docker virtual BTRFS image too, and they can't be corrected either.

It all depends on who you ask. I currently do not recommend BTRFS at all, and recommend avoiding it.

Mind explaining why?

It all depends on who you ask. I currently do not recommend BTRFS at all, and recommend avoiding it.

Mind explaining why?

 

Already documented in numerous threads across the forums. Mostly plenty of users experiencing errors at the filesystem level with no means of fixing them. The position of LT staff is if users have any issues with btrfs or other items its strictly because of their hardware. That was explored more in the numerous threads where it was definitely not hardware related.

Already documented in numerous threads across the forums.

That was explored more in the numerous threads where it was definitely not hardware related.

This is, in my opinion, one of the biggest benefits AND problems with unRAID.  There is so much information in the forum threads that those that keep up have insight that is almost impossible to find elsewhere.  At the same time people that don't have time to monitor the threads get pointed to random links, and have no idea if the "community opinion" has changed based on some later set of threads that are more recent.  And I'm sure you get tired of getting asked the same questions.  Add to this the fact that this is an opinion-based question, and it gets even tougher.  Maybe the stackoverflow model would make sense here?  /rant

 

Thanks for the info.  I'm not sure what I will do yet, but I appreciate people talking the time to reply.

And another part adding to the problem of such a vast array of knowledge is most users are using and reading the forums on mobile devices, which doesnt have the best means of copying and pasting information. On top of that, the search functionality of SMF is horrible and down right pathetic. Sorry I couldnt provide direct links, but I'm on mobile device now, but even on a PC the searching of the forums is an utter mess and practically unusable.

And these are the exact reasons LimeTech may never see implementation into the Large Scale enterprise corporate environments.

 

What's worse you now have someone like Linus of LTT talking about BTRFS in his systems, which now 100's of thousands of viewers are getting possibly misleading information

 

This is the only thing that truly frustrates / infuriates me about unRAID.  True support is not even a joke, it's less.  I really hope it's something Tom and the team plan to change.

 

Then again, who knows, maybe they like where they're at.  I just could easily see unRAID dominating the large scale corporate , and enterprise systems, if they just focused on organizing some sort of real support.  (I know it's a lot, believe me I know.)

 

/rant

Just to give some perspective, myself and many users have a cache drive formatted as BTRFS and do not have issues, and no I don't think it is a "just wait until you do" kind of thing.

 

In the past a previous installation I had (with another cache drive) and a beta version of 6.0 I did have an unmountable situation, that was worrisome, but I did have a hardware issue and some other things going on.

 

I use my 512 SSD to cache everything, with 4 VM's always in use, and recorded TV (and live tv buffer) writing to it all the time. No issues, works flawlessly..

I purchased the "Pro" Samsung drive with the 10 year warranty, or some ridiculous amount of data transfer, so I have no worries of writing to it as much as I do (that's kind of the purpose).

Anyhow to each their own, and I know there are some issues with possible data recovery (so I fully understand the other side to this), but if I replaced my SSD again, I'd still stick with BTRFS.

 

At this point in time without the snapshot features, I don't know that I'm getting any other benefits over it from XFS. I do know the scrub feature and some others could be beneficial, but I have not used them yet.

One of the selling points of BTRFS is that the filesystem stores parity information, so it can detect data errors. This was being touted as giving the ability to correct data errors by LimeTech, but the parity can only be used to detect data errors on a single disk, not correct them. I Pointed that out to Jon in the fall when he was posting how the correcting ability of BTRFS was such an advantage on a BTFRS filesystem. You need an array to have the ability to correct data errors. Linus is running a BTRFS array so even if his array has a small corruption issue running scrub should be able to fix it.

 

There is also parity on the metadata and it can be used to correct a metadata issue on a single drive.

 

 

I found BTRFS simply didn't work for me. The main place I saw issues was with my MariaDB (or MySQL) database which would become corrupted within a few days. This is why I say try it and beat on the drive with a bunch of reading and writing and then run scrub or check the log and see if it's having issues. If it works, then you're good.

 

Since I had issues, I've been running the database on XFS for almost a year without any random corruption, so I don't believe I had a hardware issue causing random corruption because it wouldn't just go away with a different filesystem. When I do the next upgrade, I will likely buy a couple of SSD's and try them in a BTRFS array. There are advantages to the filesystem if it will work.