Re: pg_clog (?) problem with VACUMM

[Gerardo Herzig <gherzig(at)fmed(dot)uba(dot)ar>]

Gregory Stark wrote:

> "Gerardo Herzig" <gherzig(at)fmed(dot)uba(dot)ar> writes:
>
>  
>
> > Hi all. Well, im having this problem for the first time.
> > When executing a VACUUM, i have this msg:
> > gse_new_version=# VACUUM ;
> > ERROR:  could not access status of transaction 118
> > DETAIL:  could not open file "pg_clog/0000": No existe el fichero o el
> > directorio
> >    
>
> []
>
>  
>
> > Ok, wtf, i will create the file. So touch pg_clog/0000 and VACUUM again....
> > ERROR:  could not access status of transaction 118
> > DETAIL:  could not read from file "pg_clog/0000" at offset 0: Conseguido
> >    
>
> That's almost the right idea for a last ditch attempt to extract what data you
> can from a corrupted table. You have to fill the file with nul bytes though.
> Something like dd if=/dev/zero of=0000 bs=1k count=nnn where nnn is, uh, I'm
> not sure how large, it won't take much to cover transactionid 118 though.
>
>  

Mmmm, yeah, after dd'ing the 0000 file, VACUUM execute just fine!!
BTW, a previous atempt to build a CLUSTER gives me the same error as 
with VACUUM, so i dont think it was a VACUUM issue.

Thanks Gregor for your help!!

Gerardo