svn (subversion) access denied for everybody

Yup. d.o SVN is totally fubar, it has nothing to do with staging.d.o per se. DamZ and I looked into it a bit yesterday. Our initial guess is version-skew for the copy of berkeley DB installed on the various webnodes, and it's screwing up the code to export infra.d.o users to the SVN user database. But, we both ran out of time before we really got to the bottom of it. Probably someone from the OSL side will have to look and fix it, since it's likely to involve changes that us non-OSL folks can't reliably make on the machines.

FYI: here's what you get when you try to read the user.db file:

% db_dump /var/webroot/infrastructure.drupal.org/svn/svnusers.db
VERSION=3
format=bytevalue
type=hash
db_pagesize=4096
HEADER=END
db_dump: page 1: illegal page type or format
db_dump: PANIC: Invalid argument
db_dump: /var/webroot/infrastructure.drupal.org/svn/svnusers.db: pgin failed for page 1
DATA=END
db_dump: PANIC: fatal region error detected; run recovery
db_dump: close: DB_RUNRECOVERY: Fatal error, run database recovery
db_dump: PANIC: fatal region error detected; run recovery
db_dump: dbenv->close: DB_RUNRECOVERY: Fatal error, run database recovery

We tried regenerating the file, and got the same thing. Seems like Berkeley DB is really confused on our boxes...

is #630834: error updating local svn copy related to this?

The issue here was the new centos webnodes not having db4 4.6. I have solved this by locking infra.d to one of the 4.3 versions and svn should be working now.

Confirmed, this is fixed. My apologies for the inconvenience .

Yup, I confirmed it's working for me, too. Thanks for the fix.

Reopening, we do have the same issue again.

Moving infrastructure.drupal.org to www5 should solve the problem.

The cron is run on the incorrect webnode, leading to the user file being scratched all the time. Could we restrict infrastructure to www5, pretty please?

I've moved the cron run to www5. Thanks for noticing that.

-N

Closing this as fixed.