Re: svnadmin recovery takes "forever" tip

Erik Huelsmann wrote:

> Please stop using subversion 0.*; non-beta versions have been released
> for the major part of last year now. They provide you with more
> stability and many bugfixes since the last 0.* release.

I understand this point of view, but have been in the happy situation
where no need was there to drive the incentive for an upgrade.

We have experienced absolutely no problems with 0.37, it supports the
features we use and have never given any problems. The server is not
publicly available, so we don't have to worry about scurity and the
dump-format lets us migrate upwards if we have a problem of any kind one
day.

I think of it as a luxury that we have a version which works for us.
That upgrades are not required, but can be chosen on the merit of their
relevance to the user. It may be beta to you, but it's rock-solid here,
with the use we have.

Someday, when a new fancy feature which we would like to use makes an
update required, or if we hit a bug, then i'll gladly upgrade.

It's a case of weighing "You know what you've got, but you don't know
what you're gonna get", "If it works, don't fix it", and so on against
the need for new features, and we havent need any of then yet.

BTW: I have a "personal" svn-server on 1.0.9, fsfs backend, but I don't
have 20 people breathing down my neck if that breaks down, so I just
upgrade that whenever a new thing I want to try out comes out.

The wedge was caused by the sysadmin stopping apache while a checkout
was going on, this would wedge the repo in recent version too, right?

I wasn't pointing fingers at or bug-reporting on 0.37. Actually, I don't
think it's very important which version of svn generated the problem, I
just included the version-number for reference. As I understand it,
recent bdb-backend versions would have the same problem? except that
dbd-logs are cleaned by default now? (actually, the logs were from even
before 0.37, which we upgraded to because it fixed a copy bug for us and
was the latest "stable" back then).

The reason for my mail was to get into the archives that you might save
hours of waiting for "svnadmin recover", if you were lucky. So that the
next unlucky soul might have an easier time than me when facing the same
problem. I searched the archives for a mail like that, and didn't find
it, so I figured it would be nice to have it there.

> PS: you can remove unused bdb logs to reduce recovery time too. see
> 'svnadmin help list-unused-dblogs'

Thanks for the tip. That was the one that would have been of value to me
earlier. Maybe I just didn't search hard enough for the info? (although
I was googling for about an hour...)

With this tip, svnadmin recover takes just a few minutes on the
afflicted repo.

BTW: Wasn't that what i did (rm log.*) "the hard way". No unused
transactions should be left when the repo is wedged, right?

-- 
Helge
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org