I have a 2TB ext4 file system (Ubuntu running Linux kernel 2.6.31-22-server x86_64). This file system is the second drive on a Drobo box plugged in via USB. We've not had problems on the first drive (Drobo limits drive size to 2TB due to some OS limitations, so if you have more space than that it appears as two separate drives).
I am sharing this files with Samba (smbd 3.4.0) with a mix of Windows and Linux workstations.
Recently we've been experiencing some data corruption in multiple files. In many cases I have an un-corrupt original file stored on one of the workstations. These are binary files of various formats, (e.g. SQLite, but others as well).
I used "split" to split a corrupt and uncorrupt file into 4096 byte chunks (this is the block size of the ext4 file system). I then ran md5sum on pairs of chunks and discovered that the chunks matched in many cases and in every case where they did not match, the corrupt chunk was a solid chunk of zeroes (
620f0b67a91f7f74151bc5be745b7110 for what it's worth).
I'm trying to track down a culprit but am a bit at a loss. I don't believe Samba is at fault since I'm using it without issue on the first drive exported by the Drobo.
What can I do to narrow this down and find out what's going on?