Okay, I think I've finally resolved the problems I've been seeing with
subversion. Long story short, it turns out it was a hard drive issue
after all.
For those interested in more detail... It turns out there's a little
incompatability/bug/weird stuff going on with some of our Seagate
Barracuda drives on certain computers/controllers. If the hard
drive's write cache is enabled, it ends up corrupting data, but only
for really large files (eg. our dump files). Some things that were
throwing me off, however, was that some of the hard drives have the
exact same model in exactly the same computer and they didn't have
these problems (probably because they were on a different controller).
I haven't been able to pinpoint it any further than that, though.
Another interesting red-herring, was that when I recompiled David
Summers' rpms on our RedHat 9 system in order to use db4-4.1.25-8
instead of compat-db-4.0.14-15, the new rpms did not have this
corruption on the dump files. (I did have to disable the
'make_*_check' options in the rpm to get it to compile, however, but I
don't know if this was due to the corruption).
So looks like even version 0.16 doesn't have non-deterministic dumps
once I disabled write-cache.
Has anyone else seen strange problems like this with Seagate drives
(or any other drives, for that matter)? I'm trying to figure out
exactly where the problem is... drive, controller, linux driver...
but I haven't figured it out yet.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Tue Feb 17 22:52:01 2004