CVS outage (and how we're going to get independent of SF)

Post by **Lucifer** » Wed May 17, 2006 1:39 am

Hmm, I've always thought of SCM as a workspace anyway, not a historical log. Sure, it has history in it, but I always figured the primary purposes of SCM were "Oh shit, I broke it, better get the old one before my boss finds out" and allowing parallel development via branches (and collaborative development from one source, of course). And history just kinda proceeds from there... Heh.

Anyway, I was looking for the How To Setup A Darcs Node For Armagetron Advanced and didn't see it.

Does anybody actually know how we'll do this, or is it one of those things we can't know until after the svn migration is done?

Website · Post by **Z-Man** » Wed May 17, 2006 7:28 am

Right. The main point of an SCM for us is the collaboration bit, that it makes sure developer A does not overwrite the changes of developer B. Then comes the branching ability. Then the history in the sense of "Shit, we found a security hole. We need to go back to version 0.2.8.2, fix it there, and release an update immediately."

And if the workspace bit is really the most important thing for you, then no matter what Luke does, you'll be able to rearrange acme in CVN to your liking. SVN is quite flexible about that. A SVN repository is essentially just one huge workspace where the history of all file/directory copy, move and change operations is kept. If you don't care about the history all that much, you can just move old and obsolete acme out of the way in a single command (svn move acme obsolete-acme-luke-imported-against-my-wish) and start fresh. Or do the rearrangements you want to do bit by bit as you would do without SCM, only with "svn move" instead of mv, "svn copy" instead of cp and "svn delete" instead of rm.

Still, it's your call what should happen with it during the import. I'd hold that it's enough that the all so important history is preserved in CVS.

For the Darcs repository, we'll need to write small conversion scripts. I have full featured ones, with tags and branches, for CVS, in a private CVS repository. I'll rewrite them for SVN, probably at first without respecting tags because that's too difficult there, and publish them when they're ready. Small control files that say "create a darcs repository for all files below this point" will have to be added at strategic points of the SVN repository, that's for sure.

Post by **Tank Program** » Fri May 19, 2006 6:23 am

Back on the topic of the distributed MySQL database I think I've worked out a way to make that work with the forums and I'll test that later today.

Post by **Lucifer** » Tue May 23, 2006 8:20 am

Here's another one to throw on the pile, stumbled across while fooling with my car computer project.

http://www.rocklinux.org/wiki/SubMaster

Website · Post by **Z-Man** » Tue May 23, 2006 9:55 am

Looks like the SubMaster client does almost the same as SVK, but it seems to have tamer dependencies. I *almost* got it to work on my PC here

(SVN::Mirror test phase fails) Everyone is free to use it for personal disconnected operation as long as we stick to a writable SVN repository.
Now, the SubMaster server is interesting. For those too lazy to read, it's a web pased patch management system, a bit like the "patches" tracker on SourceForge, that also helps with applying and testing patches. The SubMaster client helps with generating patches the right way. Together, they something similar to "darcs send" and "darcs apply", only managed on the Web. Nice find.

Website · Post by **Z-Man** » Tue May 23, 2006 2:21 pm

Looking closer, you can't use the SubMaster client to conveniently send your collected changes to the svn server. You can only create patches or use "svn commit" itself, which then works like SubMaster wasn't used at all and does a single commit. That means it is no replacement for SVK. You can't go back in history with a SubMaster checkout, but that's secondary.

Post by **Lucifer** » Thu Jun 01, 2006 8:44 pm

Apparently sourceforge doesn't even want to give us time to set something up.

I don't have my mirror of svn going yet.

Website · Post by **Z-Man** » Thu Jun 01, 2006 10:05 pm

I have one, but it's not yet accessible via http(s) and syncs once a day only anyway. But hey, the rsync server is still up.

Post by **Lucifer** » Fri Jun 23, 2006 3:37 am

Ok, I've got my svn backup script running now. It's running hourly, so if there's another outage I'll at least have a pretty up-to-date backup of the repository. Whether we do anything with it or not, I don't know.

wrtlprnft · Post by **wrtlprnft** » Fri Jun 23, 2006 4:09 am

You could link it to armabot and only run it on every commit

Luke-Jr · Post by **Luke-Jr** » Sat Jun 24, 2006 2:44 am

Anyone know a good redundant network filesystem that supports file locking (more-or-less network RAID-1)? Such should allow us to at least in theory move the Svn repository to the network filesystem and run multiple redundant Svn hosts all supporting read-write.

Post by **Tank Program** » Sat Jun 24, 2006 7:44 am

Might Coda support it?

joda.bot · Post by **joda.bot** » Sat Jun 24, 2006 10:12 am

I read it supports replication...
http://en.wikipedia.org/wiki/Andrew_file_system

It's used at our university to offer access to windows and linux homes.

Post by **Lucifer** » Sun Jul 16, 2006 5:47 am

Tank Program wrote:Back on the topic of the distributed MySQL database I think I've worked out a way to make that work with the forums and I'll test that later today.

I seem to recall this test running and being successful. What's next? Now that armabot will run commands that let us run certain updates automatically when svn information changes we have the possibility to put the forum software in svn, and if you make sure your changes are committed, we can have the mirrors update the code all at the same time (relatively). Same with the main site.

I guess we still need to run a few more tests of various sorts, and my test drupal instance is currently down (part of the gentoo move, I'll get it back up if we want it, but my drupal assignment has dropped off and I'm back in mamboland for awhile, so drupal hacking is off the radar unless at least one other person here volunteers to worry about it with me).

Post by **Tank Program** » Sun Jul 16, 2006 9:13 pm

I've worked out a pretty much 100% system for replicating the forums, minus doing some of the actual work. All the proofs of concept are in place.