naturally, i can no longer reproduce it after reloading the recompiled module. I'll keep using the patched module and hopefully the problem resurfaces. I did find out however that one of my 3 storage array's was spitting out occasional corrupted data. I had to copy files using rsync over 20 times to finally get them to match. (rsync -cPa [A-G]* /dest to copy in 'blocks' and always made sure to be copying bigger chunks then memory available). What's odd is that I have ecc ram, edac enabled memory controller, but no errors whatsoever anywhere to be found. after a raid5 'check' command, nothing wrong. Yet copying 2.7 TB from 1 array to the other worked flawlessly, yet copying 20gb from the 3rd to either of those produced lots of errors.
The purpose of this explanation totally unrelated to aufs is, emerge --sync was syncing to a directory on this broken raid5 array. I have thus decided to migrate everything off of there and remove that array. Thus if the error keeps emerging later on again, it wasn't the broken array, if not, it was. I did run the find / -ls thing on the array but found nothing interesting. Just a 'no such file or directory' error on files that where there. I'm now guessing I may have forgotten to umount my aufs'd tree while copying/removing some of the files on the underlaying FS, but not sure now. On 02/09/11 15:54, sf...@users.sourceforge.net wrote: > Oliver Schinagl: >> How do they smell buggy? I've used pmfs for months now, albeit with the >> wrong mount options (see previous post). I'll try the patch and >> hard-reboot my machiene again :S > Because I have fixed around pmfs few bugs which means I forget most of > pmfs code and behaviour. > If you can, try these steps. > - find /your/rsync/source /your/rsync/destination -ls > lists > - strace -o trace -f rsync ... > > These "lists" and "trace" will be important to see where the problem > happned. > If your system totally stopped and you cannot retrieve files after > crash, then it is better to store the files remotely via network, serial > or anything else. But it is totally up to your system environment. > > > J. R. Okajima ------------------------------------------------------------------------------ The ultimate all-in-one performance toolkit: Intel(R) Parallel Studio XE: Pinpoint memory and threading errors before they happen. Find and fix more than 250 security defects in the development cycle. Locate bottlenecks in serial and parallel code that limit performance. http://p.sf.net/sfu/intel-dev2devfeb
