For everyone who pulled during the 30 minutes the extra certs were on venge.net... (was Re: [Monotone-devel] Re: lots of new certificates?)

[Nathaniel Smith]

On Wed, Mar 29, 2006 at 12:36:24AM -0800, Nathaniel Smith wrote:
> On Wed, Mar 29, 2006 at 09:54:38AM +0200, Lapo Luchini wrote:
> > Timothy Brownawell wrote:
> > > A number of old revisions now have two sets of certs, one by
> > > address@hidden (the certs you already had) and one by
> > > address@hidden (the new certs).
> > Ah ok, I feared it was some kind of bug in which the certs didn't
> > "match" no more and were re-doundloaded ;-)
> > 
> > BTW: isn't there a functionality like "dropping many certs based on a
> > pattern matching"?
> > (well, thinking about it, I think it does, it's called something like:
> > mtn db execute 'DELETE FROM <certs> WHERE <match>')
> 
> Yeah, we're working out exactly what happened and how to get rid of
> them on IRC right now, I'll post a recipe once finished...

Right, so, here's the appropriate recipe:
  $ mtn -d <your db> db execute \
    "delete from revision_certs where keypair = 'address@hidden'"

It's been run on the server, if you pulled the extra certs, please
make sure you run this command before sync'ing again.  (Hopefully
everyone who did pull will see this in time; if not, maybe we'll have
to bump up the priority on that cert allergy thing we've mentioned a
few times to deal with situations like these... :-).)

If you're not sure whether you need to run the command, you can just
run it; it's safe.

An easy way to tell if it worked, is to run 'db info' before and after
the command, and see if the number of certs in your db changes by
~10,000.  I had 30,000-something certs in my db after pulling, and
then back to 20,000-something after running the above command.

Cheers,
-- Nathaniel

-- 
"...All of this suggests that if we wished to find a modern-day model
for British and American speech of the late eighteenth century, we could
probably do no better than Yosemite Sam."