FSCK-I/O errors

[Chet Woods]

Hi all,

I'm having problems with my /var filesytem - /dev/vg00/lvol9. I'm trying to run an fsck and it won't work. Here's the error - fsck: filesystem had I/O error(s) on metadata. I ran a dd and the disk as a whole seems ok. My problem however is this...obviously that is part of the OS, and I do not have a back up of it (well I do have an old make_recovery). This is a remote server and we just cancelled hardware support on these servers - OLD E25(due to budget constraints, and I'm not happy about it).

I'd really like to find some other resolution than building another server ans shipping it up there. If anyone can help I sure would appreciate it!!

Fyi - I already tried the fsck -o full option, didn't work. Also, getting it in single user mode (and asking the end user to do the typing) is available but, not recommended.


thanks everyone,
Chet

[Michael Steele_2]

I didn't see reboot anywhere.

'fsck' will automagically run during boot up.

Why can't you single user? Cause it's an E25?

reboot.

[A. Clay Stephenson]

Well, when you choose not to make backups regularly and not to have mirrors, this is the price you pay. Since you didn't specifically mention it (and judging from your attempts I assume that you know this is, in fact, a vxfs filesystem) the very first thing to try is another fsck -F vxfs -o full,nolog -y (note the nolog).

If all else fails, you should be able to build a new /var filesystem and then extract just the /var filesystem from your make_tape_recovery but obviously that would be a dated copy but far better than nothing.

[Chet Woods]

We can get into single-user mode however, it's somewhat risky. At that point I'd be asking a "non-technical" user to do the typing at the console and read to me any errors. I'm rebooting at this point and we'll see what happens. If a reboot doesn't clear it up I'll try what Clay suggested.

thanks for the quick response.

Chet

[Chet Woods]

Clay,

Obviously I know that not backing up or not mirroring isn't best practice. I'm just an administrator that doesn't makes the rules, I can only give my opinion to my bosses. It's a sad, less than recommended cost saving world I live in here. :-(
I'm going to try and build a new /var filesystem and attempt to recover from my make_recovery (yes, not even make_tape_recovery) tape. Any suggestions before I dig myself deeper in this hole?

Thanks,
Chet

[A. Clay Stephenson]

There are several threads that deal with doing partial extractions of a make_tape_recovery:

Here is but one:
http://forums1.itrc.hp.com/service/forums/parseCurl.do?CURL=%2Fcm%2FQuestionAnswer%2F1%2C%2C0x2aa844f56197d711abdc0090277a778c%2C00.html&admit=716493758+1066768241600+28353475

Before making a new filesystem, the very first thing that I would do is practice on your local box using a make_tape_recovery so that you can get the commands exactly right when you email/fax/phone them to the assistant. This will also get you familiar with the commands and the data displayed so that you know what to expect even when possibly translated into dingbat-ese (excuse me, user jargon).


The newfs, partial make_tape_recovery extraction is probably your least evil option.

[Michael Steele_2]

If you get up to run level 3 run this command:

# check_patches (* attach errors from /tmp/verify.log *)