[Pushmi-users] Issues with HTTP svn access

[Chia-liang Kao]

Hi,

> I've successfully got it installed (I think? CPAN + CentOS 5 != a good
> time) but am hitting some issues that I can't find good answers for:

Just a thought, did you run the test suites for SVK and Pushmi?

Also, what svn version are you running for master and slave?

> === issue 1 ===
> 
> When attempting to sync the master to the slave when the master is using
> HTTP I get a sync failure without fail:
> 
> lwebb at hamsambo /tmp/SUP-569/bin $ ./pushmi mirror
> /tmp/SUP-569/slave_http http://127.0.0.1/svn/repos/rd
> Mirror initialized.
> lwebb at hamsambo /tmp/SUP-569/bin $ ./pushmi sync --nowait
> /tmp/SUP-569/slave_http
> [/tmp/SUP-569/slave_http] locked by /tmp/SUP-569/slave_http:svm:lock:_
> Retrieving log information from 1 to 17250
> [/tmp/SUP-569/slave_http] unlock result: 1
> [/tmp/SUP-569/slave_http] sync failed: RA layer request failed: REPORT
> request failed on '/svn/repos/rd/!svn/bc/17250': REPORT of
> '/svn/repos/rd/!svn/bc/17250': 200 OK (http://127.0.0.1) at
> /tmp/SUP-569/lib/perl5/site_perl/5.8.8/SVK/Mirror/Backend/SVNRa.pm line
> 469

Do you have any authz configured on the repository?

> Apache is reporting the following:
> 
> lwebb at hamsambo /tmp/SUP-569/bin $ sudo tail -n 1
> /var/log/httpd/error_log
> [Tue Jul 10 15:47:23 2007] [error] [client 127.0.0.1] Provider
> encountered an error while streaming a REPORT response.  [400, #0]
> 
> lwebb at hamsambo /tmp/SUP-569/bin $ sudo tail -n 1
> /var/log/httpd/access_log
> 127.0.0.1 - scmbuilder [10/Jul/2007:15:46:45 +1000] "REPORT
> /svn/repos/rd/!svn/bc/17250 HTTP/1.1" 200 2609609 "-" "SVN/1.4.2
> (r22196) neon/0.25.5"
> 
> When the same master repository is accessed via file:///path/to/master
> instead of http:// the sync appears to work properly
> 
> I notice that there is some references to the sync failures with SVK,
> but none appear to be associated with an apache code of 200.
> 
> Is there anyway that I can extract some debug logging from either SVK or
> Pushmi to assist me in working out what is at fault here?

If you can run wireshark and get a dump of the conversation over
localhost it'd be quite helpful.

> === issue 2 ===
> 
> When the sync is working it is terribly slow without much noticeable
> load on the master or the slave.
> The master repo has 17325 revisions within it & after some 5 hrs of
> running the slave is only at revision 6055.
> 
> I'd be happy to blame the hardware but the machine has 2 x Dual-Core AMD
> Opteron @ ~ 2Ghz, & is able to svnadmin load the master repo in about 20
> minutes or less
> 
> Is there anything I can do to analyse the poor performance?
> 
> Cheers in advance for any assistance.

Again, a network dump would be helpful to see if it's latency-bound.
Meanwhile, please check if pushmi (or svk) became two processes while
syncing, if not somehow pipelining is disabled and has some performance
impact.

Cheers,
CLK