Missing reports in GNATS

Missing reports in GNATS

[Wolfgang Bangerth]

The "disk full" problem of yesterday seems to have been fixed, but now 
there are two reports missing. You can see that when you ask GNATS for 
_all_ reports: it (presently) shows you "8987 matches found", but the 
highest one has number 8989. One of the two missing ones is actually my 
favorite, PR/38 ;-) I don't know which is the second.

Is it possible to restore these?

Thanks
  Wolfgang

-------------------------------------------------------------------------
Wolfgang Bangerth              email:           bangerth@ticam.utexas.edu
                               www: http://www.ticam.utexas.edu/~bangerth

Re: Missing reports in GNATS

[Angela Marie Thomas]

> The "disk full" problem of yesterday seems to have been fixed, but now 
> there are two reports missing. You can see that when you ask GNATS for 
> _all_ reports: it (presently) shows you "8987 matches found", but the 
> highest one has number 8989. One of the two missing ones is actually my 
> favorite, PR/38 ;-) I don't know which is the second.
> 
> Is it possible to restore these?

I cannot find 8988 or 8989 in the backups.  The highest number I have is 8987.

Looking at the gnats administration files, the highest number it has is
8987 as well.  However, it is the case that c++/38 and pending/8975 are
not in the gnats index file.  If someone remembers how to manually lock
the index file I can fix it.  If someone knows how to regenerate the
index file, that would be even better.

--Angela

Re: Missing reports in GNATS

[Tom Tromey]

Angela> If someone knows how to regenerate the index file, that would
Angela> be even better.

See the `gen-index' program.
It is in /usr/local/libexec/gnats/
Docs are in info, but from memory, run as gnatsid:

gen-index -n -d /sourceware/gnats/gcc-db -o /sourceware/gnats/gcc-db/gnats-adm/index

Tom

Re: Missing reports in GNATS

[Angela Marie Thomas]

> Angela> If someone knows how to regenerate the index file, that would
> Angela> be even better.
> 
> See the `gen-index' program.
> It is in /usr/local/libexec/gnats/
> Docs are in info, but from memory, run as gnatsid:
> 
> gen-index -n -d /sourceware/gnats/gcc-db -o /sourceware/gnats/gcc-db/gnats-ad
m/index

Thanks.  I'm updating the index now (writing to a .new file to test it
out first).

--Angela

Re: Missing reports in GNATS

[Wolfgang Bangerth]

On Wed, 18 Dec 2002, Angela Marie Thomas wrote:

> > The "disk full" problem of yesterday seems to have been fixed, but now 
> > there are two reports missing. You can see that when you ask GNATS for 
> > _all_ reports: it (presently) shows you "8987 matches found", but the 
> > highest one has number 8989. One of the two missing ones is actually my 
> > favorite, PR/38 ;-) I don't know which is the second.
> > 
> > Is it possible to restore these?
> 
> I cannot find 8988 or 8989 in the backups.  The highest number I have is 8987.

8988 is the first one that was added today, so if the backup runs 
overnight, then it will not have it.

W.

-------------------------------------------------------------------------
Wolfgang Bangerth              email:           bangerth@ticam.utexas.edu
                               www: http://www.ticam.utexas.edu/~bangerth

Re: Missing reports in GNATS

[Angela Marie Thomas]

> 8988 is the first one that was added today, so if the backup runs 
> overnight, then it will not have it.

Yes, the backups run in the wee hours of the morning US/Pacific.

Hopefully it's just an issue of regenerating the index file.  It doesn't
look like gen-index does any locking, so I'm first seeing how long it
takes.  I suspect we will need to shutdown gnatsd in order to regenerate
the index correctly, but would like to know for how long before requesting
that.

--Angela

Re: Missing reports in GNATS

[Wolfgang Bangerth]

On Wed, 18 Dec 2002, Angela Marie Thomas wrote:

> > 8988 is the first one that was added today, so if the backup runs 
> > overnight, then it will not have it.
> 
> Yes, the backups run in the wee hours of the morning US/Pacific.
> 
> Hopefully it's just an issue of regenerating the index file.  It doesn't
> look like gen-index does any locking, so I'm first seeing how long it
> takes.  I suspect we will need to shutdown gnatsd in order to regenerate
> the index correctly, but would like to know for how long before requesting
> that.

Given the amount of traffic that goes through GNATS, I would be more 
comfortable if you shut it down, so as to prevent inconsistencies. Even if 
it takes some time. If you do it in the later afternoon PST, you also 
won't block someone from Europe, as they should be sleeping by then.

Thanks for taking care of this
W.

-------------------------------------------------------------------------
Wolfgang Bangerth              email:           bangerth@ticam.utexas.edu
                               www: http://www.ticam.utexas.edu/~bangerth

Re: Missing reports in GNATS

[Angela Marie Thomas]

> Given the amount of traffic that goes through GNATS, I would be more 
> comfortable if you shut it down, so as to prevent inconsistencies. Even if 
> it takes some time. If you do it in the later afternoon PST, you also 
> won't block someone from Europe, as they should be sleeping by then.
> 
> Thanks for taking care of this
> W.

By taking advantage of the disk cache and doing some sanity checking,
I was able to update the index file "live".  I ran the index generator
a few times to preload the disk cache, did a diff of the current index
with the new one to ensure the only changes were the two missing PRs
and did a double check of the gcc-prs list.  Not an ideal process, but
one for which I have high confidence.  The last message in gcc-prs
matches the date on the old index file so I'm reasonably sure we're
up-to-date now.

Finding a "good" time for doing this sort of thing is next to impossible
which is why I opted to go with this solution.  Normally I would rather
just shut it down as well, but was unsure whether anyone would be free
later to deal with it (I'll be in meetings) and didn't want it forgotten.
Also, I wanted to ensure things were up-to-date before the next backups.

--Angela

Re: Missing reports in GNATS

[Wolfgang Bangerth]

> matches the date on the old index file so I'm reasonably sure we're
> up-to-date now.

Thanks, Angela!

W.

-------------------------------------------------------------------------
Wolfgang Bangerth              email:           bangerth@ticam.utexas.edu
                               www: http://www.ticam.utexas.edu/~bangerth