Subversion, FishEye/Crucible (the Source + Review Bundle) on OnDemand End of Service

[cofarrell]

Hi Henry,

Would it be problematic at all to complete fully the migration of Subversion and only then move on to the migration of Fisheye/Crucible?

Not at all. The two are really separate operations, so migrating SVN first shouldn't cause any problems.

Also, most references show Unix/Linux command line statements when it comes to installation and mantenance of the tools, if installing these tools on an Windows system would the command line statements be the same?

The command line should be almost identical for Unix/Linux and Windows. We support both, and we don't necessarily recommend one over the other. My suggestion would be to pick the system that you have the best support for and understanding of.

Could one maybe install TortoiseSVN on the Windows server and perform the create repository, load repository, and other maintenance operations through that GUI instead of via command line?

Anything command that is svnadmin I'm fairly certain (although please correct me if I'm wrong) you'll have to do on the command line. Anything that's just 'svn' should be possible with TortoiseSVN, although it probably isn't worth the trouble once you've gotten that far in the migration (it's just testing the repo layout).

Good luck.

Charles

[Norman Abramovitz]

I am just verifying there will no way to use On-Demand products with any subversion respository after October 2013. If that statement is true, is there anything on the Altassian roadmap to allow subversion repositories to interact with On-Demand?

[Henry Widman]

Another question Helen.

If one had the choice of installing Subversion and Fisheye/Crucible on either a Unix/Linus or Windows server (as a virtual system) would there be a preference? Is either system in anyway better than the other (e.g. installation, ease of setting up and maintaning user access, performance, general maintenance)?

Also, most references show Unix/Linux command line statements when it comes to installation and mantenance of the tools, if installing these tools on an Windows system would the command line statements be the same? Could one maybe install TortoiseSVN on the Windows server and perform the create repository, load repository, and other maintenance operations through that GUI instead of via command line?

Thanks,

Henry

[Henry Widman]

Our plan is to migrate Subversion and Fisheye/Crucible to a local instance. The migration instructions say to backup and download the Subversion repository and then start the migration of Fisheye/Crucible, and then to continue the migration of Subversion. Would it be problematic at all to complete fully the migration of Subversion and only then move on to the migration of Fisheye/Crucible? Our critical path is to get Subversion migrated and working (as time and resources are limited), the Fisheye/Crucible migration is important to us but not the priority.

Thanks,

Henry

[aMarcus]

Bitbucket Support is available through support.atlassian.com under the Bitbucket Support project. The public issue tracker is for tracking bugs and feature requests with the site itself, similar to jira.atlassian.com for the other Atlassian products.

[Steve Muskiewicz]

Here's another migration related question - how do we receive support for Bitbucket issues given that we are a paying OnDemand customer. I see that Bitbucket has its own issues tracker in https://bitbucket.org/dashboard/issues but this doesn't seem to be very formal.

Can/do we use the support.atlassian.com portal to file issues with bugs we find in Bitbucket?

thanks!

-steve

[Joni]

Help! In the last migration phase it says:

3. Create an account for your build tool on Bitbucket

Create an account for your build tool (Bamboo OnDemand) so it can read the Git repository on Bitbucket. The best practice is to grant read-only access to this build account.

How do I do that? :-o

[cofarrell]

Hi Joni,

You should be able to create a normal Bitbucket account and then grant that user read-acccess to your repository. You would then supply Bamboo with the crentials or private SSH key of that user.

Another option, which isn't mentioned, is to use deployment keys, which has the advantage of not adding to your plan limits.

Does that help?

Cheers,

Charles

[Joni]

Hi, I tried to create a deployment key, but I received a 'protocol not supported' error. https://answers.atlassian.com/questions/90338/using-a-private-key-to-connect-to-a-git-repo-securely-with-ssh-in-bamboo-on-demand

[cofarrell]

Hi Joni,

That's most unfortunate. :( You should definitely vote for that Bamboo issue.

For now you will have to create a separate user, or use your own credentials. The latter is quicker/easier, but isn't generally a good idea for production builds as they will break if you change your password.

Charles

[Chad Gilmer]

Patrick,

So then, we can use JIRA+Confluence and Bamboo with Bitbucket by just "configuring" it right ? we probably need the repository migrated 100%. Are we able to use the "source" tab in the issues when using Bitbucket ?

In my test without any integration with JIRA, keys for the projects are ignored, I supposed for commits you will have the same constraints like when using SVN for commiting code,etc. So it works like before except for the code, that will be in a GIT repository.

Thanks for the link, I probably read that already. It's very general.

Is that all the technical documentation that you have regarding the topic ?

Kind regards

[Kinto Soft]

It looks like there are a lot of people happy migrating from Subversion to Git... good luck for you!... but maybe not all from them are so happier:

It's a pitty that Atlassian, few days ago, announced that 3rd party V2 plug-ins are not longer supported for onDemand, because there are coming alternatives to hosting Fisheye in order to support Subversion on JIRA. The following is one of them:

https://marketplace.atlassian.com/plugins/pbeltranl.subversion.jira.pbeltranl-jira-subversion-plugin

As mentioned, the door has been closed (and locked), but within few months I hope to release a V3 compatible edition of this plug-in with much more premium features like graphical support for merge and reports... and it should run on onDemand once V3 plug-ins are supported and this is ready.

So, if you are not happy being forced to migrate to Git, be a bit patient...

[Mauricio Leyzaola]

Hi, at first we thought about moving to Mercurial but as other people have pointed, we are by no means experts and fear thing can go wrong with our code.

So, as of today we have setup a linux server with Subversion Edge and during these months we are going to migrate our projects from FishEye to our own server.

I would like to use that same SVN server to host FishEye and this way, we shouldn't miss any of the current features we have in Jira (source linking, reviews, etc). Is it possible for FishEye to live on the same machine with SVN? If it is, could you let me know how?

Thanks

[Patrick Hill]

Mauricio,

Many of our customers have Subversion and FishEye on the same server, this is also how we run FishEye + Subversion in OnDemand. As a starting point, have a read through of our FishEye Documentation (https://confluence.atlassian.com/display/FISHEYE/Subversion) to see how to set up your Subversion repository with FishEye.

Patrick

[Joe Parrilla PEGS OLD]

New Issue!

We have about 16 projects, all of which I have successfully cloned. This last one however is giving me trouble. After cloning for about 3 hours (the project is the largest), I get this message:

Following parent with do_switch
perl: /build/buildd/subversion-1.7.5/subversion/libsvn_subr/dirent_uri.c:1519: uri_skip_ancestor: Assertion `svn_uri_is_canonical(child_uri, ((void *)0))' failed.
error: git-svn died of signal 6

I read a bunch about it, mostly having to do with URI's not being the right format. I went into the .config file to make sure there were no double slases or trailing slashes, which there werent. The URI's seemed to be in the correct format. None of my other projects had any issues, only this one. Is there any explanation as to why this would happen? What project specific issue could cause this that would not be present in any of my other 16 clones?

[Chad Gilmer]

Hi,

I have a simple question to do, I started to migrate some SVN repositories to GIT for testing, and automated it so I do not have to go one by one necessarily. Now I have all commite changes,etc. In new repositories, so migration from SVN to GIT seems to be possible with not too much trouble.

The problem is, what is gonna happen with the Jira wiki (confluence), Bamboo and the tracking system (issues) in Jira and the Bitbucket repository ?

Now we have FishEye integrated with issues, and every project in Jira has source code and issues associated, and the wiki. I see BitBucket has wiki and tracking system too... but we do not want to loose all this data.

We want to use Jira and just use BitBucket for the source code using GIT, is that possible ? is that the idea in the migration ?

I really need some explanation on how to integrate Confluence and Jira with BItBucket.

Kind regards

[Patrick Hill]

Chad,

You've 100% correct! Bitbucket is ready to be swapped out directly for FishEye/Crucible to host your code, along with reviewing code amongst team members. The wiki and issues in Bitbucket can be used for smaller projects, however as an existing JIRA + Confluence user you are able to integrate Bitbucket to track your code. As a starting point, have a read through https://confluence.atlassian.com/display/AOD/Getting+started+with+Bitbucket+and+OnDemand to get an idea of how Bitbucket works with OnDemand.

Patrick

[Joe Parrilla PEGS OLD]

@Peter

@Michael

Thanks. I was able to use a 3rd party downloader (downthemall) in Firefox. This allowed the entire file to complete downloading. Thanks for the advice. I just could not figure out what was happening.

Next problem... So after doing the load local-svn < svn.dump portion of the migration, I am not ending up with an actual local project structure. I was under the impression that my local-svn directory should contain a structure such as local-svn/PROJECT1, local-svn/PROJECT2, etc. Instead the bulk of my folder is contained in a file called db/revs. There are no projects visible, just folders named 0, 1, 2, 3, etc. I am trying to create the author mapping file, but no project directories exist! The size is accurate (~3.2 GB), but there are no actual project directories such as trunk, branches, tag, etc.

[hhung]

@Steve Muskiewicz

Yes, you're right. Unfortunately, the Source tab does not currently link to Bitbucket - here's the relevant ticket: https://jira.atlassian.com/browse/DCON-288

[Michael Knight]

When downloading in the browser, the download "completes" at only 1 GB and I end up with a corrupt svn.dump.gz file. Is this an issue on my end or yours?

This is a problem at our end. See AOD-5975 for details and workarounds in the comments.

Could a team member commiting during the dump cause corruption in the dump file? I was under the impression that commits were not allowed while the dump is being generated.

You can commit during the export generation process but it will not corrupt the repo or the dump. The dump will only contain commits that were present before the export started. You may want to adjust permissions before doing the export to lock it down beforehand.

I now cannot create another dump for 48 hours, which puts a huge hold on the process.

You can contact Support and we can help you out.

[Peter Drier]

@Joe.. yep, more bad documentation from Atlassian.. To get the file > 1gb, you need to use a "wget" like command with a continue or retry option.. The file is there and is valid, it just takes longer than the allowed web timeout to get the whole thing..

try something like:

wget -r --tries=10 http://foo.jira.com/dump/svn.dump.gz

and it should work

[Joe Parrilla PEGS OLD]

I created the subversion dump file, which is claimed to be ~3.3 GB. When downloading in the browser, the download "completes" at only 1 GB and I end up with a corrupt svn.dump.gz file. Is this an issue on my end or yours? Could a team member commiting during the dump cause corruption in the dump file? I was under the impression that commits were not allowed while the dump is being generated. I now cannot create another dump for 48 hours, which puts a huge hold on the process. Thanks.

[Steve Muskiewicz]

We've been doing some testing with Bitbucket and OnDemand and the integration seems to be fairly good. However one thing that we noticed today is that the "source" tab in a JIRA issue is not linking back to the commit/source changes in Bitbucket (even though the "Commits" and the "Activity" tab both seem to have Bitbucket linkage there)

This might be a source of confusion for some of our developers who are used to looking at this tab. Is this something that is supposed to work or is that tab only for linkage to Fisheye?

Similarly will the "Reviews" tab have any linkage to pull requests in Bitbucket?

thanks!

[Paul O'Flynn]

How can I tell how big my SVN repo is? I searched the online doc but can't seem to find an answer.

Thx.

[Patrick Hill]

Paul, You can download the SVN Dump file from the backup manager to get an idea of how big the repository is.

[Paul O'Flynn]

How can I tell how big my SVN repo is?

Thx.

[collegescheduler]

Couple of questions, I'm going to run Criucbile/Fisheye myself on my own server.

1. If I only have 4 active Crucible/FishEye users but I have had 10 users in the past, what license do I need? In other words, I've got a ton of Crucible/FishEye history for checkins, comments in Crucible, etc. but those other 6 users don't work for me any more. Can I get away with the 5 user license instead of having to buy 10 seats? (which is a HUGE price difference.) I don't want to lose any past data, checkins, comments, etc.

2. I know that you are offering a free 12 months of the Crucible license... does that 12 months start from the date I switch to hosting my own Crucible? I know we have about 5 months left to switch, but I don't want to switch today and lose those 5 months of free service keeping with my OnDemand.

[hhung]

Hi Robert

1. A user in Crucible is defined as anyone with a "Can Use" permission (http://www.atlassian.com/software/crucible/pricing) so if you only need 4 of these then there's no problem with the Starter license of 5 users. The history will still be kept, those old users will not have access to the system though.

2. That's correct, for FishEye or Crucible download, it's 12 months from when you go through the prescribed switching process.

[Peter Drier]

@Helen, I'd agree with you if your original instructions were 100% vetted and accurate.. but they're not.. I listed 3 things above which they were (and still are, WHY HAVE YOU NOT UPDATED THEM???) wrong or missing. Thus comments are necessary.. They would only get comments of people having problems that need to be fixed.. and others would help.. that's the way this kind of world works.. but instead Atlassian seems to be against it's own users lately.. This whole migration is harder than it needed to be because of the poor documentation and not well though out process. Too many manual steps that could have been completely automated.

[hhung]

Hi Peter - yes, you're right and I'm sorry about this. We're looking into improving the feedback mechanism.

Meanwhile, we have taken your feedback on the guides seriously and after some discussion with the team today, we realised that we need a bit more clarity on the points that you raised. I created a support ticket on your behalf and hopefully you'll be able to help us out on that there.

[Nanda Kumar]

We have done a migration about a month back there were about 35 svn projects and about 30 users. All went well but the only glitch was the real time sync that is mentioned. As peter has mentioned what the sync go as far as as the local repository that you restore from backup. It does not sync from the actual JIRA SVN. We ended up doing the entire migration during a weekend and swtiching over to GIT-Bitbucket from the preceeding Monday. We had setup a test GIT project on Bitbucket to educate our users. And did a sample migration of some of the projects and documented it. On the migration weekend the Bitbucket was completely wipedout and redid the migration from our doc. Now all is going well.

[Colin]

I agree, extremely risky migration for many of us - I think Atlassian have badly mis-calculated on this one

[Peter Drier]

@Paul, I got an e-mail about your comment but it seems gone

To answer.. this seems the only place to asking migration questions.. all of the other pages telling us how to migrate (and there are multiple with differing instructions) don't allow comments.. Atlassian is losing my respect.

[Paul O'Flynn]

For such a large (and potentially dangerous) undertaking I'm very disappointed that this the only place to ask qs :-( The migration guide is a nice start in my mind but that's all it is really. I've used git in my last job, but never atlassian and never svn but I got dumped with this migration project and will definitely be asking qs... Oh well... Thx for your responses Peter.

[hhung]

Hi Peter and Paul - if you have migration questions, we indicate on our migration website that you can raise a support request. We'll be able to help you out as much as possible: https://support.atlassian.com/secure/CreateIssue.jspa?pid=10840&issuetype=50

Otherwise, yes we decided to make use of this Answers thread so that all public q/a and comments can be consolidated in the one spot for ease and everyone's benefit.