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

Corrupted repository report

From: Scott Lawrence <slawrence_at_pingtel.com>
Date: 2004-06-17 18:20:28 CEST

The "book" says that if we have a corrupted repository that 'svnadmin
recover' won't, we should save it and report it... so here goes...

Running svn server 1.0.3, with all repository write access via apache
2.0.48 (there is no svn: access, and all file: access is read-only
(backups and viewcvs). Platform is RedHat Linux ES3 (svn installed
using the summersoft RPMs, thank you David). Client is svn command line
client, version 1.0.3 (r9775) compiled May 24 2004, 10:48:57 on gentoo.

The repository is new; it had only 4 revisions in it so far. 222 files,
2201463 bytes of file content, 33 directories. We had had some trouble
when importing the files; the 'svn import' reported:

Adding sipviewer/bin
Adding sipviewer/bin/sipviewer.in
Adding sipviewer/Makefile.am
Adding sipviewer/build.xml
svn: MERGE request failed on '/rep/sipXtackLib'
svn: MERGE of '/rep/sipXtackLib': timed out waiting for server (https://scm.sipfoundry.org)

This appears to have been a result of a slow pre-commit hook that was
installed in the repository. After removing the hook for the import,
the files were imported successfully, and an 'svnadmin verify' on the
server at that time (revision 1) reported that all was well. We
reinstalled the hook, and did a few additional checkins of smaller
numbers of files successfully (up to rev 4). After I left for the
evening, the developer tried another larger checkin (modifications) and
hit the timeout again. When I saw that report I took hook out again
(and will leave it out pending a rewrite - all it does is check for tab
characters in program sources; can live without it). Our nightly backup
(based on 'svnadmin dump') ran successfully a few hours after that.

This morning, the developer retried the checkin and got:

Transmitting file data .........................svn: Commit failed (details follow):
Berkeley DB error while committing Berkeley DB transaction for filesystem /scm/rep/sipXtackLib/db:
Invalid argument

I ran 'svnadmin verify' on the server and got:

svn: Berkeley DB error while checkpointing after Berkeley DB transaction for filesystem sipXtackLib/db:
Invalid argument
svn: Berkeley DB error while closing 'nodes' database for filesystem sipXtackLib/db:
Invalid argument

so I ran 'svnadmin recover' on it and got:

Please wait; recovering the repository may take some time...
svn: DB_RUNRECOVERY: Fatal error, run database recovery

I moved the repository, recreated it, and reloaded it from the backup;
the restored repository seems to be working (sans the hook); this
mornings problem checkin worked on the restored repository.

The bad repository can be downloaded from:

  Scott Lawrence
  SIPfoundry server tools maintainer
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Thu Jun 17 18:21:53 2004

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.