Thread: Maintaining 2 SVN Repositories

Hi all,
I have two SVN repositories that I would like to mirror, one that is public facing, hosted remotely; and the other for my private use hosted locally (in case the public one should ever go down). For every commit I need to have it commited to both repositories. I have not yet found a non-frustrating way to do this; I've tried:

a) committing changes to one then copying over changed files and commiting to the other (and thereby commiting myself to an insane asylum)

b) committing changes to my private repository then every so many changes doing an svn dump then load into the public repository. And with it dumping about 80% of my bandwidth cap and 5+ hours of upload time, not to mention the public database is rarely current.

c) I tried dumping just the diffs; however this neither worked, or if it did would be just as frustrating having to dump-load every change made

What is a solution, possibly scripted, that I could use to commit a single change to two repositories? Or in other words, how do I maintain two identical SVN repositories without going insane?

Or is SVN the wrong tool for the job?

I'm sure you could write little wrappers around 'ci' etc, which basically do the same command twice over.

> And with it dumping about 80% of my bandwidth cap and 5+ hours of upload time
How much historic junk do you have in there?

Originally Posted by Salem

Or is SVN the wrong tool for the job?

Yeah, I had a similiar desire myself and switched to the use of a distributed version control system.

SVN is pretty much not able to do this (easily,) by design. What you really want is the ability to 'offline commit', meaning you can just make commits that stay in your repository and then you push them publically when you want.

There is a tool, SVK, which allows you to do this in the context of SVN (specifically, it is an SVN client with offline commit and other stuff.) Note: it will *not* make it easier to commit to two different SVN servers, it will merely allow you to pool commits locally until you decide to push them to the actual, user-viewable server.

I however, recommend jumping ship from SVN and using a distributed version control system like Mercurial or Git. They allow you to work exactly how you want and offer much more.

In terms of "historic" I have everything since 1; however all the junk is filtered out. Is that not normal?

Thanks for the suggestions all, I will look into Git (which is distributed version control?). For those who have made the change, how much of a hassle is it to adapt to Git coming from SVN?

Originally Posted by @nthony

Thanks for the suggestions all, I will look into Git (which is distributed version control?). For those who have made the change, how much of a hassle is it to adapt to Git coming from SVN?

I am not a git user, but when I switched to using Bazaar, I found it quite easy to adapt to the similiar command set. Of course, things get more complex if you want to move further into full utilisation of DVCS, but you can always progress in steps.

I have heard comments that git is likely to be harder to switch from Subversion than Bazaar and Mercurial, but then git advocates also claim that after you achieve proficiency it is superior (but then all software tools more or less say the same thing, heh).

Note that all three of the DVCS mentioned have some support for working with Subversion repositories (and working copies), which helps in a gradual conversion. I believe that Bazaar has the best support with respect to Subversion, but I could be mistaken.

Originally Posted by laserlight

Note that all three of the DVCS mentioned have some support for working with Subversion repositories (and working copies), which helps in a gradual conversion. I believe that Bazaar has the best support with respect to Subversion, but I could be mistaken.

Thanks, this will be important. Also glad to discover that SourceForge supports all VCSs mentioned in this topic!

You can do this with SVN easily.

Write an SVN hook. Use `svnsync` to synchronise the repositories. The end.

Originally Posted by zacs7

Write an SVN hook. Use `svnsync` to synchronise the repositories. The end.

not sure why the hook is necessary, but I think it works, thanks... I was just starting to learn Git

I highly recommend trying git, it's fantastic. Much like you, I was working on a SVN-managed project and I wanted something similar (offline commits), and many people suggested I try a distributed VCS. I picked git (I'm sure bazaar and many others are great too), migrated my SVN repository over to git with a simple one-line command (I forget exactly what, just google it), and that was it. All SVN history was incorporated into git with no issues.

It takes some time to learn a new version control system, and git's learning curve is probably higher than a simpler system like CVS, but it's well worth it in the end in my opinion . . . .