Ben writes:
> You say that "it could be a bad bug", but it's not clear what you mean
> by "it". What's the problem? You showed us a DB error about not being
> able to open the 'nodes' table, followed by a successful 'svnadmin
> recover'. Is the error now fixed? Does it persist? We need a better
> description of what's wrong.
Sorry for the inadequate details. I was trying to balance
being overly verbose with being helpful with enough details.
I had snipped a bunch of verbose output from a couple runs
of my "recover svn1 in case I can't be reached" script that
I've provided for our ops department.
I'm doubly sorry for the inadequate details because as
I was composing the email with adequate details, I found
the problem.
Really Short answer:
Some bug happened, which recovery fixes. No particular
complaint there.
Longer answer:
My initial svnadmin recovery was run as part of a script
that "does it all" (stops httpd, recovers repos, starts
httpd and checks on the health of the repositories by
hitting their URLs). That script would successfully
stop/recover/start, but the main repo remained wedged.
This was because the script was still recoving the 0.34
repositories, not the 0.36 repositories we're actually
using. (They have the same URL, but different locations
on the disk)
Once I ran recover on the *CORRECT* repo, it did in fact
work, making this *CLEARLY* not a "bad bug". (I had
thought that even-post-recovery the repo stayed wedged,
which would have been bad.)
Kudos on a great product, and apologies for the initial
report.
--Jim
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sun Feb 1 18:28:24 2004