Re: file system errors?

[Wouter Verhelst <w@uter.be>]

Hi Christian,

On Fri, Nov 10, 2017 at 01:50:55PM +0000, Schöne, Christian wrote:
> Hi,
> 
> I’m having some issues running NBD on a LAN that seems to have some packet
> loss.
> 
> Clients are mounting a BTRFS volume RW, the server has COW (copy on write)
> activated, so all changes are going to a diff file.
> 
> Sporadically I see clients doing this:
> 
> kernel: BTRFS error (device nbd0): bad tree block start 1600888930041856
> 216154112
> 
> kernel: BTRFS error (device nbd0): bad tree block start 1600888930041856
> 216154112
> 
> kernel: BTRFS: error (device nbd0) in btrfs_run_delayed_refs:3009: errno=-5 IO
> failure
> 
> kernel: BTRFS info (device nbd0): forced readonly
> 
> Mounting originally like this:
> 
> /dev/nbd0  /  btrfs  rw,noatime,discard,compress=lzo  0 0
> 
> NBD client and server seem not to have any problems at this point. At least
> they are not logging anything. From my understanding TCP should take care of
> transmission problems. I don’t understand why btrfs is complaining. NBD should
> be reliable here, right?

It should be, yes.

> Is btrfs maybe not the best choice for nbd? Is it possible it has some issues
> with higher latencies?

Honestly, I'm not all that familiar with btrfs.

> Is there a best practice file system that is known to work nicely with NBD?

I've not heard of many issues with most filesystems that could be pinned on
NBD. Did you try any others?

Which kernel are you using?

-- 
Could you people please use IRC like normal people?!?

  -- Amaya Rodrigo Sastre, trying to quiet down the buzz in the DebConf 2008
     Hacklab