Christian Vogt wrote:
> Hi,
>
> we had a crash yesterday destroying the logs from subversion (1.1.3)
> database. I've dumped the dabase tables (db_dump), created a new
> repsoitory and loaded them (db_load). Via svnadmin load/unload failed
> with "panic ...." messages.
> So far so good, all seems to be nice after loading.
> Now I'm trying "svnadmin verify" and get an checksum error (expected x,
> found y) direct for revision 0. Also performing a "svnadmin recover"
> doesn't help.
> My users getting more and more errors and we are nearby rev 8000 (round
> about 7,5 GB).
>
> Is there any solution to force recalculation or so?
Really depends what the extent of the corruption is. A checksum of all null
bytes will be considered as a match-anything wildcard. Problem is that if
the corrupted data is a directory listing, then subversion is just going to
error again when it reads the corrupt directory entry, and if it's in some
file data, it's probably going to make some earlier versions unrecoverable
too, since their delta base will be corrupt.
Under normal circumstances, I'd suggest that you put a tarball of the broken
repository somewhere for me to analyse, however 7.5 GB is rather large.
Max.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Sun Mar 13 01:47:42 2005