On 14-Sep-04 Roland Hill wrote:

Hi there,

I wonder if anyone could give me a few pointers? Running SuSE 8.2Pro.

I had a power cut last night while composing an email in mutt. When I brought the server back up I had the folloing issues:

- A lot of file permissions changed. Generally to root.root. This crashed postfix. I found all/some of the problems and postfix started. I noticed some of 'my' files had also changed in the same way. [and other woes ... ]

A good friend reminded me about running fschk over the drive (didn't do it last night due to mild panaic attack <g>), so I will do this tonight. All ideas are most welcome.

'fsck' may help. It's a remarkably capable program and can solve a surprising range of filesystem problems, considering the state that file systems can get into. However, you are likely to find that a lot of files (and file fragments) end up in the "lost+found" directory of the relevant filesystem. Nevertheless, I'm sadly inclined to suspect that you have suffered major corruption. A simple power cut should not change permissions of files etc. -- the computer would normally simply stop, in the middle of whatever it was doing when the power failed. As a result, the worst you should expect after re-boot is that disk-writes from cache were not carried out, so some file changes did not take place and there may possibly be some inconsistencies between the contents of different files; and "hidden" backup files created during editing are still there, etc. That sort of thing. At worst you should find when 'fsck' is run automatically (since not shut down cleanly) is that some inode deletions need cleaning up, and inode counts resetting, all of which 'fsck' should take in its stride. Changed permissions is much more ominous. This suggests RAM failure. Maybe (the more likely) they are being wrongly read from disk, so that it looks like "root.root" but in fact is not (remember that userids are stored as numbers in inodes, not as names). Root has user number "0", so all the bits in the user number are zeroed. Possibly (though unlikely because it seems several files are involved) this took place during a disk-write from corrupt RAM. Or -- worse still -- your hard drive circuitry has been damaged. I would suggest some hardware tests before trying to put things right, just in case the cure turns out to be worse than the disease. Eg: 1. Run "memcheck" on your RAM. This is available as a boot-time option so you do not actually involve your Linux system in this. 2. Physically take out each hard drive, and install it in some other machine where you can mount its filesystems in the Linux system running on that machine. What does the BIOS see? What do you get when you run 'fdisk /dev/hdWhatever' on the drive? (only use the "p" command in fdisk, to print the partition information; DON'T TOUCH the "w" key because this will write stuff back to the partition table. Finally quit with "q".) Then check the contents of directories, ownerships, permissions, etc. This enables you to separate hardware problems like corrupt RAM from problems like damaged hard drive circuitry. 3. If you have a spare old hard drive, install it in the suspect machine (with no other hard drives present) and try to carry out an installation of Linux on this. Run various tasks on this for a bit and check whether anything unwanted is happening. This can check the motherboard/hdd interface as well as whether the CPU/RAM are doing their job properly. Unfortunately the power surge that sometimes accompanies a power cut and/or its restoration can cause such hardware damage. Living myself where brief power cuts are all too common, I have my two desktops running off a Belkin uninterruptable power supply (UPS). Belkin provide Linux drivers for their UPSs, and it works well. There is a serial link from the UPS to one machine which invokes a "shutdown" script if the power has been off more than a certain time, and the UPS then waits for a further time before switching itself off. Meanwhile machine 1 (from this "shutdown" script) sends an email over the LAN to user "shutdown" on machine 2, which has a root 'cron' job watching /usr/spool/mail/shutdown/, and if a file appears there then machine 2 removes it and does "shutdown -h now". After sending the email, machine 1 shuts down. There's just time for all this to happen before the UPS goes off. Result: any power failure lasting more than a minute or so leads to clean shutdown of both machines, therefore no file corrpution. Brief power cuts (up to 10 secnds at a time are frequent) are simply ignored: the UPS continues to supply power, and all is back to normal once power resumes. The UPS procides an excellent buffer against surges and spikes coming up the mains, and I have had no problems since starting to use it. On the other hand, I thought I was OK with a laptop in this situation, since a laptop has its own battery and can ride out even quite extended power outages, and can be set up to go into 'shutdown' when a "battery low" condition develops. However, I was rudely woken out of this complacency one day when the power went off/on/off/on/... several times in quick succcession. After this, the laptop was dead. The subsequent repair attempt found that the motherboard had been fried. The replacement motherboard being of the wrong type, all this machine is now good for is an archive of files from which, when needed, I can recover copies by splitting tar archives into floppy-sized chunks (yes, the PCMCIA slot doesn't work either so I can't download over the LAN!). Now I use a replacement laptop which is isolated from the mains through two layers of Belkin "SurgeMaster" surge protector. Beware! Good luck with your attempts at recovery. Best wishes, Ted. -------------------------------------------------------------------- E-Mail: (Ted Harding) Fax-to-email: +44 (0)870 167 1972 Date: 15-Sep-04 Time: 16:08:04 ------------------------------ XFMail ------------------------------