[Xapian-discuss] Compressed Btrees (bugged?)

[Arjen van der Meijden]

So I was finally able to have quartzcompact run like I wanted it.
But after it had finished, omega didn't want to search through the 
database anymore, it gave 'inflate err = -3' as output and aborted.

When quartzcheck ran over the database-files it yielded this same error 
on the postlist-table. Other tables are, afaik, not tested on the same 
extensive level so the results from quartzcheck on those tables is 
probably less usefull? The b-trees of those tables are all ok.

Anyway, it seems to me that either the read- and write-processes are not 
equal or something went wrong at either? How can I find out about that?

Best regards,

Arjen van der Meijden

On 14-12-2004 11:20, Arjen van der Meijden wrote:
> It turns out that the position-table took much more than 3 hours to 
> compress using the zlib-layer with minimal compression results. I don't 
> know the exact times, but I killed it after it had run for 3:38u while 
> it had only done 4.7G of the data. An earlier non-zlib run took a bit 
> over 1:15u on the same machine with a similar load.
> 
> So for the next runs I won't be compressing that table anymore. The 
> termlist took only 3 minutes more orso to generate.
> 
> Best regards,
> 
> Arjen
>