Why ReiserFS?

[darrenscerri]

Why did unRAID developers choose ReiserFS as the used file-system? From what I read, ReiserFS is somewhat slower than ext3 for large files (which is what unRAID is mostly used for), and also very slow in mounting.

 

Good read: link

[WeeboTech]

From what I remember of prior conversations.

 

1. So that inodes do not have to be pre allocated and users do not have to be concerned with having free space and not enough inodes.

 

2. At the time of unRAID's initial design, reiserfs was the only filesystem that supported safe easy expansion.

[sacretagent]

well i wanted to ask this question already a few times ...

 

why does reiserfs do such a terrible job at finding and repairing(moving) bad blocks...

do you guys have some program that does this automatically or you just trow away a disk the moment you have sectors pending ?

 

would another filesystem like ZFS or even ext3 not be easier to manage small disk issues ?

 

 

[WeeboTech]

I don't know that reiserfs does such a terrible job at find/repairing.

There are two parts here.

 

finding/repairing bad blocks. That's at a lower level and I think you need to use the badblocks tool to get down to that level.

I'm unfamiliar with ext3 being able to diagnose those situations.

 

Now at repairing the higher level. We've seen people do disasterous things to their file systems and reiserfsck was able to recover many parts of file systems that I believe ext3 would never have been able to deal with.

[Chris Pollard]

or you just throw away a disk the moment you have sectors pending ?

 

 

Personally I do this.  If its under warranty then it goes back to the manufacturer,  if not it goes on ebay with a smart report in the description.

[limetech]

From what I remember of prior conversations.

 

1. So that inodes do not have to be pre allocated and users do not have to be concerned with having free space and not enough inodes.

 

2. At the time of unRAID's initial design, reiserfs was the only filesystem that supported safe easy expansion.

 

That's right.  Also, that article mentions very long mount times; in my experience that has not been the case.  The "mkfs" operation is also much faster than was ext2 at the time.  In developing the unraid driver I added code that looked at how "sequential" block allocation was for large files and I remember that reiserfs was hands-down better at "large block" allocation - that is, keeping large file data close together.  Anyway those tests were a few years ago.

 

One other thing worth mentioning: I have found reiserfs to be pretty resilient.  There have been cases where people have started a parity sync where their "parity" drive was actually a data drive and then realizing this aborted the parity sync - and then being able to recover a lot of data.

 

I know there is not a lot of new development with reiserfs (v3) - getting rid of the "big kernel lock" dependency was probably the last real "big change".  And yes, the original developer is in jail - I guess that's one way to stop feature creep   But it has proven to be a good fit with what it's used for.

 

I think the next best candidate would be xfs.

[aiden]

I think the next best candidate would be xfs.

 

I know you're not opening this up to discussion about the alphabet soup of everyone's favorite file systems, and why they are superior to the others.  But with ZFS having so many benefits for file protection, is it ruled out because it's 1) is a Linux port, and therefore not native to the kernal, and 2) doesn't allow for online expansion?  And are the benefits of ZFS generally mitigated by the potential for P+Q parity in unRAID?

[SSD]

I don't know that reiserfs does such a terrible job at find/repairing.

There are two parts here.

 

finding/repairing bad blocks. That's at a lower level and I think you need to use the badblocks tool to get down to that level.

I'm unfamiliar with ext3 being able to diagnose those situations.

 

Now at repairing the higher level. We've seen people do disasterous things to their file systems and reiserfsck was able to recover many parts of file systems that I believe ext3 would never have been able to deal with.

 

I agree wholeheartedly with this.  It is simply amazing how well RFS (reiserfsck) has been able to reconstruct a disk after various drive calamities.

[sacretagent]

but again..

is there a tool that automatically repairs reiserfs filesystem with bad blocks ?

now i need to start reiserfsck -- check

wait till it finds a bad block

 

dd this bad block manually

 

and restart reiserfsck...

 

been busy now for 6 hours like this and it gets ennoying ...

i know the drive is for the garbage but i just want to get one file from it ...

running unraid without a parity for a while since i needed the space and can not add a drive for the moment ...

 

windows ntfs remaps bad blocks automatically .. and if you really have a bad block then chkdsk /r will emap that one for ya .. without having to restart the proggie 20 times...

 

[neilt0]

I think the next best candidate would be xfs.

 

I accidentally deleted 50GB of data from one of my NAS devices running xfs yesterday and found that there were no unerase utilities available for it.

[WeeboTech]

I don't know that reiserfs does such a terrible job at find/repairing.

There are two parts here.

 

finding/repairing bad blocks. That's at a lower level and I think you need to use the badblocks tool to get down to that level.

I'm unfamiliar with ext3 being able to diagnose those situations.

 

Now at repairing the higher level. We've seen people do disasterous things to their file systems and reiserfsck was able to recover many parts of file systems that I believe ext3 would never have been able to deal with.

 

I agree wholeheartedly with this.  It is simply amazing how well RFS (reiserfsck) has been able to reconstruct a disk after various drive calamities.

 

After seeing some of these disasters, then seeing people come back with a significant portion of data, I did not need to have ext2/3/4 on the array any more. (Although I could really see having ext3 or 4 for an apps drive that I may want to swap between machines).

[WeeboTech]

but again..

is there a tool that automatically repairs reiserfs filesystem with bad blocks ?

now i need to start reiserfsck -- check

wait till it finds a bad block

 

dd this bad block manually

 

and restart reiserfsck...

 

been busy now for 6 hours like this and it gets ennoying ...

i know the drive is for the garbage but i just want to get one file from it ...

running unraid without a parity for a while since i needed the space and can not add a drive for the moment ...

 

windows ntfs remaps bad blocks automatically .. and if you really have a bad block then chkdsk /r will emap that one for ya .. without having to restart the proggie 20 times...

 

 

The badblocks program does this at the lower drive level. not at the filesystem level.

Reminds me of a spinrite check.