I have a PCI USB 2 card, which I use to plug in external USB 2 hard disks for backup. I have one external disk that I’ve been using for months now, and it’s been working fine. Some months ago I bought a second drive, with the intention of swapping between the two, and keeping one in my desk at work as an off-site backup. It worked for a while, but then started failing during backups. I’d lost the reciept, so I ended up selling it locally to somebody under the understanding that it may fail, and if they don’t like it they can return it.
The day before yesterday, I bought a new drive to try again. I stuck it in and backed up about 200Gb to it. Then I simulated 7 nightly backups all without incident. Last night, our power glitched for about 3 seconds, which was long enough to knock both the external drives off, because they’re not on the UPS. This morning I power cycled them and mounted them both, and did another nightly backup. But in the middle of backing up to the second drive, it started failing in a very similar way to the old drive. It started with:
Mar 15 07:51:30 allhats kernel: lost page write due to I/O error on sdb1
Mar 15 07:51:30 allhats kernel: sd 18:0:0:0: rejecting I/O to device being removed
Mar 15 07:51:30 allhats last message repeated 47 times
Mar 15 07:51:30 allhats kernel: sd 18:0:0:0: rejed 18:0:0:0: rejecting I/O to device be
ing removed
Mar 15 07:51:30 allhats kernel: sd 18:0:0:0: rejecting I/O to device being removed
Mar 15 07:51:30 allhats last message repeated 1154 times
Mar 15 07:51:30 allhats kernel: __journal_remove_journal_head: freeing b_frozen_data
Mar 15 07:51:30 allhats kernel: EXT3-fs error (device sdb1): ext3_get_inode_loc: unable
to read inode block - inode=7897505, block=15794191
Mar 15 07:51:30 allhats kernel: journal commit I/O error
Two external hard drives, both of which caused these sorts of problems under load. Hmmm. I’m starting to wonder if it’s not the disks. Maybe it’s something about the USB controller. I need to test this somehow. I wonder if my work computer has USB 2?
Mar 15 07:51:30 allhats kernel: lost page write due to I/O error on sdb1
Mar 15 07:51:30 allhats kernel: sd 18:0:0:0: rejecting I/O to device being removed
This doesn’t look it’s a drive issue; it looks like the USB side has decided the device has been removed (unplugged) while still mounted. My machine seems to have a dodgy connector and if I move the cable a little it can happen. The disk goes away and then comes back as a new device (because sda1 was still mounted, it comes back as sdb). Annoying.