Re: [rdiff-backup-users] no space left on regression

[The Anarcat]

dean gaudet wrote:
> On Mon, 15 May 2006, The Anarcat wrote:
> 
>> /: write failed, file system is full
> ...
>> IOError: [Errno 28] No space left on device
>>
>> Why the heck does it go write in /? /tmp? Why not use a tmp dir in the
>> target? And how can it use 130M of tmp?
> 
> that first error looks more like some general libc or other such thing 
> giving you the filesystem the error occured on rather than a specific 
> directory, but i dunno.  strace would tell you for sure.

Well, it does tell me it's in /, so I suspect it's the partition. The
backups are done in /backup2, a seperate filesystem with plenty (ie.
100GB+) of space.

/ has much less left (130M), but that's where /tmp sits.

> anyhow, rdiff-backup needs temp space to regress files because it needs to 
> apply the reverse deltas... and it would be unsafe to do that in-place in 
> the originals (i'm not sure librsync even supports doing it in-place).  i 
> thought in this case it actually did use temp space on the target 
> itself... but it's been a long time since i looked at that code.  
> searching the mailing list archives for tmp might get you some answers.

Right, okay.

> you're going to need to give it space somehow... maybe you'll get lucky 
> just by deleting the target/rdiff-backup-data/file_statistics.* files ...

Bah, i'll just create a /tmp partition. :)

> if not you might try aging the oldest backup... if that doesn't work you 
> might end up having to delete it by hand (by figuring out the timestamp 
> which will be embedded in every related filename and deleting those 
> files/directories from target/rdiff-backup-data).

... aging the oldest backup? What does that mean?

A.

signature.asc
Description: OpenPGP digital signature