From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.5) Gecko/20041111 Firefox/1.0.2 Description of problem: If anaconda would use smartctl to check a HDD for errors before installation proceeds, this would save some people some grief. (On the other hand, I'm glad Fedora runs smartd by default!) Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1. Install Fedora on bad hard drive. 2. Boot. Actual Results: I was disappointed I wasted time installing Fedora on a hard drive that was failing. Expected Results: Anaconda could have used smartctl to check the drive's status. It would have discovered several "pre-fail" and "old age" flags. Additional info: After I installed and then rebooted, I got an email from smartd about these errors: Device: /dev/hda, 54 Currently unreadable (pending) sectors Device: /dev/hda, 34 Offline uncorrectable sectors smartctl -H /dev/hda reported PASS, although smartctl -a /dev/hda reported several "pre-fail" warnings.
it reported PASS because it did pass. your drive has uncorrectable errors which can be remapped, though it's a bit involved[1]. it would only FAIL if the drive wasn't able to remap them or had some other problem. [2] [1] http://smartmontools.sourceforge.net/BadBlockHowTo.txt [2] http://smartmontools.sourceforge.net/ see "My ATA drive is failing its self-tests, but its SMART health status is 'PASS'. What's going on?"
I like this suggestion for anaconda. I plan to work on this post-FC5 since to do it correctly, I don't really have enough time to get it done for FC5. smartmontools consists of userspace tools and to do this correctly, I'd prefer a library with Python bindings. Still, it's a cool idea and expect to see some activity on this in rawhide after FC5.
are there any known buggy disks/IDE controllers which would cause problems for smart checks?
Absolutely, which is why work for this needs to begin in smartmontools. I haven't sat down to look at the smartmontools code closely, but I know that a lot of controllers provide issues as well as certain manufacturers' implementations of SMART. Their FAQ mentions the well-known problem disks and controllers.
so installing and running smartmon by default as it is done in FC right now is not wise? a RFE should be opened to request smartmon NOT be installed and enabled by default... at least until smartmontools are fixed
I don't think I explained that correctly. smartmontools works well and handles a wide range of disks and controllers. It knows about a lot of manufacturer specifics. What I was trying to point out is that we simply can't rely on a pass/fail test for anaconda. We'll probably have to look at a variety of SMART data fields and decide from there. So this isn't that smartmontools is bad, it's interpreting the information it gives you in the report(s) programatically that's more difficult.
I don't think this is a feature we can realistically add to the installation process. We already have enough mechanisms that point out bad hardware in log files.