[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

large revision causes problems

From: Jan Evert van Grootheest <j.grootheest_at_euronext.nl>
Date: 2003-09-11 12:14:39 CEST

Hello,

I am having these problems with SubVersion due to a really large
revision (or two) that's in the repo.

Here's the history:
There's a tree of files that I once tried to commit. It started to
transfer the data (printing all those dots), but after a long time it
reported an error. I think the server committed everything just fine,
because the tree is there when I browse through it with a web browser.
However, the client reported an error. Unfortunately, the exact error
got lost. The WC directory tree for the commit totals some 230Mb.
The working copy that the committed tree is in, still thinks that the
add was not committed.
If I try 'svn status', it reports in 18832 lines things that need to be
added.
If I try 'svn status -u' I can see from apache that it does a lot of
things for a while and then one worker shoots up to 99.x% CPU and stays
there for a long time. At some point in time, the client reports a timeout:
[issysd25@a05008 current]$ svn st -u
svn: RA layer request failed
svn: REPORT request failed on '/svn/MAL/!svn/vcc/default'
svn: REPORT of '/svn/MAL/!svn/vcc/default': Error reading chunked
response body (http://a05009:65030)
All the while, the server is still very busy. These are the relevant
lines from the apache access log:
a05008.aex.nl - - [10/Sep/2003:16:53:30 +0200] "PROPFIND
/svn/MAL/MAL/Dev/documentation/3rdparty/ACE/current HTTP/1.1" 207 758
a05008.aex.nl - - [10/Sep/2003:16:53:30 +0200] "PROPFIND
/svn/MAL/!svn/vcc/default HTTP/1.1" 207 397
a05008.aex.nl - - [10/Sep/2003:16:53:30 +0200] "PROPFIND
/svn/MAL/!svn/bln/16 HTTP/1.1" 207 450
a05008.aex.nl - - [10/Sep/2003:16:53:30 +0200] "PROPFIND
/svn/MAL/!svn/bc/16/MAL/Dev/documentation/3rdparty/ACE/current HTTP/1.1"
207 769
a05008.aex.nl - - [10/Sep/2003:16:53:30 +0200] "PROPFIND
/svn/MAL/MAL/Dev/documentation/3rdparty/ACE/current HTTP/1.1" 207 758
a05008.aex.nl - - [10/Sep/2003:16:53:32 +0200] "PROPFIND
/svn/MAL/!svn/bc/16/MAL/Dev/documentation/3rdparty/ACE/current/ACE_wrappers
HTTP/1.1" 207 2802
a05008.aex.nl - - [10/Sep/2003:16:53:32 +0200] "PROPFIND
/svn/MAL/!svn/bc/16/MAL/Dev/documentation/3rdparty/ACE/current/ACE_wrappers/html
HTTP/1.1" 207 11436
a05008.aex.nl - - [10/Sep/2003:16:53:33 +0200] "PROPFIND
/svn/MAL/!svn/bc/16/MAL/Dev/documentation/3rdparty/ACE/current/ACE_wrappers/html/tao
HTTP/1.1" 207 6158346
a05008.aex.nl - - [10/Sep/2003:16:53:30 +0200] "REPORT
/svn/MAL/!svn/vcc/default HTTP/1.1" 200 9027147
Although I cannot guarantee that all these lines are the result of that
svn status call, the paths in them all relate it. And the number 16 (if
it is a revision number) matches that really big commit that failed
client-side that I mentioned above. As you can see, it returns some 15Mb
in total in the last two lines.

Any comments up till now are welcome.
As for questions....
What can I do to make this 'svn status -u' complete (apart from a faster
server, of course)?
Will all 'svn status -u' calls in the future have this problem?

What can I do to get this working better? How can I help you to improve
subversion with regard to this problem? I will not be able to do
extensive debugging, although investing a limited amount of time is no
problem.
I really would like to improve subversion and not expose the rest of the
company to this, as there are already some that are not convinced of the
advantages of svn over source safe (from a project point of view, not
necessarily technical).
Since, until now, there is no proprietary code in the repo yet, I could
mail/make available the database or a dump of it for testing. The db is
some 374Mb, while the zipped dump is still close to 60Mb.
Processing the dump somewhat reports these numbers:
[switch@a05009 dump]$ wc -l MAL.20030911
7258432 MAL.20030911
[switch@a05009 dump]$ grep -na Revision-number MAL.20030911
5:Revision-number: 0
15:Revision-number: 1
285:Revision-number: 2
1939:Revision-number: 3
1966:Revision-number: 4
2585:Revision-number: 5
3204:Revision-number: 6
3806:Revision-number: 7
4407:Revision-number: 8
5042:Revision-number: 9
4743620:Revision-number: 10
4743645:Revision-number: 11
4743670:Revision-number: 12
4743702:Revision-number: 13
4743724:Revision-number: 14
4746521:Revision-number: 15
4746558:Revision-number: 16

Thanks,
Jan Evert

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Thu Sep 11 12:16:08 2003

This is an archived mail posted to the Subversion Users mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.