6

I wonder if anyone can help me. We have recently migrated our Subversion repository from a VisualSVN Server on Windows to a subversion server on CentOS.

The migration was succesfull however we are getting the following error message 

Error REPORT of svn'/svn/MangoRepository/!svn/vcc/default': Could not read chunk size:
Error connection was closed by server (http://servername)

Now the workaround for this is simply to perform a update on the repo and it will contine where is left off.

Im just wondering if anyone was a permanent fix for this as it can be quite frustrating to repeat my self to 60-70 developers.

bahrep
  • 687
  • 1
  • 9
  • 27
Brian Mitchell
  • 311
  • 4
  • 13
  • 1
    Getting the message, when exactly? If the repository URL changed, users should relocate their working copies or checkout new ones. BTW see the possible dupe http://stackoverflow.com/questions/772894/updating-from-svn-repository-returns-could-not-read-chunk-size-error – bahrep Aug 04 '12 at 07:50
  • 1
    Can you test whether the error occurs with the command line `svn` client? That will narrow down whether the problem is the Tortoise or the server. And can you provide info on how you've configured the server to serve the SVN repo? – Shane Madden Aug 04 '12 at 20:00
  • 2
    Check the logs on the server. See if it says anything about why it closed the connection. – dsh Aug 05 '12 at 01:59
  • I get the message when performing a checkout, i have tried checking out using tortoise and svn client on the server itself and it comes up with the same error at the same place. – Brian Mitchell Aug 13 '12 at 10:54
  • So we using CentOS 6.3 with apache, apache is configured to authenticate against active directory using kerberos. I have checked the logs and it shows the following error "Error writing base64 data: Software caused connection abort". – Brian Mitchell Aug 13 '12 at 14:44
  • On top of this i peformed a svn dump on the repo and i set up a temp windows server with visual svn, i then loaded the same repo into this server and tried a checkout and im getting no error messages. So it seems to me that there is something in the CentOS, Apache configuration that is dropping the tcp connection. i have tired looking at help pages to try and optimize apache but im having no luck. has anyone come across anything like this before. – Brian Mitchell Aug 15 '12 at 09:32

2 Answers2

1

I've seen a similar error message when dealing with migrating from a case-insensitive filesystem (windows) to a case sensitive one (Most Unixes). Is there a chance that you have a file called Default in some contexts and default in others?

Tim Howland
  • 4,728
  • 2
  • 27
  • 21
  • The default in the URL isn't a file. It's a special URL for Subversion. So I don't think this is the case. – Ben Reser Jan 08 '14 at 17:06
0

SVN checkout send requests and receive a lot of tiny files. So it may create hundreds or thousands connections at the same time.

We have trouble with it in previous time, this is what we did:

  • Reduce the network bandwidth to decrease number of connection in a short time.
  • Adjust the firewall accept more connections.
Community
  • 1
truongtx
  • 50
  • 4