I run several servers out of my home office. I do routine backups of these every 3-4 days, at most once a week, using rsync, onto a USB hard drive (each contains some 20-30GB of data, making other backup methods impractical).

However, when server uptime gets to about 70-90 days (that is purely empirical), the USB HDD is no longer recognised (nor any other USB devices) - ie no output in syslog/messages when plugged in and lsusb does not list any attached devices.

Putting the machines into runlevel 1/S (and/or back into runlevel 3), rmmod'ing and re- insmod'ing the all the USB modules, does not solve this. Seemingly, the only solution is to reboot.

This is not /really/ a problem because usually the USB HDD is attached to my laptop and the backup is done remotely (rsync over ssh).

It's tough to reproduce this because of the the long(ish) uptimes (>~70 days)

It's just vexing.

Anybody have experiences like this or a possible solution?

Yuletide greetings, solstice salutations and peace to all slackers and LQ users.

I have no real experience in this, but still I would like to offer my . First, are you still on the 2.4 series kernels, or are you running the 2.6 , perhaps 2.6 has resolved your issue with USB, I can't say that it is a power saving issue, since why the hell would it save power after 70 days. Perhaps if you try skimming through your logs, for the past day or so, maybe the 68th to the 70th day, but yea I know that can be really daunting, but unfortunately, that might be the only thing you can do, is to try and trace back your logs. Since shutting off and turning the drive back on doesn't help, perhaps you might just want to also use a more conventional drive, like IDE or SCSI anyways, but that might seem like a last resort, but then again, I never heard anyone serving with a USB hard drive, until now, sure I know it can be done, but perhaps USB is just not optimum for your needs, where are SCSI or even IDE is. Or it may be, and it just might be a kernel issue, that is if you are running the 2.4 series. Sorry this may not be very helpful, but yea, since its hard to provide detailed information, like snips from logs, this seems like the only best answer at the moment.

Jeebiz, I think you really misread my post and your c2 are misguided and somewhat condescending. I'm not using the USB drive to serve data, I'm using it for backups. It's more convienient than tapes, and fits nicely in the bottom of my firesafe. If you had read my post, you would have easily gleaned this information.

However:

The servers are all x86 machines, some with 2.4 kernels, others 2.6. The problem seems to be consistant across all the servers - that at around 70-90 days, when I come to do a backup, I plug in the USB HDD and the system behaves as if I hadn't.

Usually I'd issue mount -t reiserfs /dev/sda1 /mnt/usb and proceed to do the backup. After a long uptime, the machine will balk: /dev/sda1 is not a valid block device. Nothing in the logs, and the drive does not show up on lsusb. It's not that I find log files daunting, it's just that there's nothing there!

As I previously stated, this is not a 'problem', but it is something that makes me go "Hmm..."