id: cons "/usr/sbin/getty" console console vt100

[BrianT_1]

On aan AlphaServer running Digital Unix V4.0F, I'm seeing the following:

INIT: command is respawning too rapidly. Check for possible errors
id: cons "/usr/sbin/getty" concole console vt100

id: lat02 "/usr/sbin/getty" /dev/tty02 console vt100

id: lat03 "/usr/sbin/getty" /dev/tty03 console vt100

id: lat04 "/usr/sbin/getty" /dev/tty04 console vt100

id: lat05 "/usr/sbin/getty" /dev/tty05 console vt100

The server will not continue beyond this. This occurred after a crash that happened when I removed a huge messages file from /var/adm created by syslogd. I changed the syslog.conf file to put the that log into syslog.dated instead and tried to restart syslogd, but got the crash and this started happening. Recommendations, please.

[Raj D.]

Brian,
Looks like you have boot in maintenance mode , Single user mode :

While the system is booting , and wait at console mode: Press Ctrl+c ,
- The go to single user mode using :
# boot -fl s

Revert back the changes, keep only syslog.conf , aso verify the /etc/inittab file .

-Reboot it back , & check.


Cheers,
Raj.

[Martin Moore]

You'd get this effect if /usr was failing to mount. Can you boot to single-user mode and run "bcheckrc" to see if /usr mounts successfully?

Martin

[BrianT_1]

I'll try both suggestions. Thanks.

[BrianT_1]

RajD wrote:

> While the system is booting , and wait at
> console mode: Press Ctrl+c ,
> - The go to single user mode using :
> # boot -fl s
>
> Revert back the changes, keep only
> syslog.conf , aso verify the /etc/inittab
> file .

WHat do I do after I get the single user prompt? I user the boot command above, the system booted to single user mode, but I can't do anything except a cat or ls command. How do I get the file systems mounted? I can't find my documentation on how to proceed in single user mode. I can't change any files without any commands.

[BrianT_1]

The errors I've posted are a result of an earlier error in the bootstrap. When mount_advfs tries to run, the following error occurs:

ADVFS: Domain usr_domain not activated - inconsistency detected
usr_domain#usr on /usr: Device does not contain a valid ADVFS file system

Since /usr doesn't mount, the /usr/sbin commands won't run and the messages I first posted occur.

I can't figure out what happened to the advfs file system. Is there a tool I can use to recover it?

[Steven Schweda]

> Is there a tool I can use to recover it?

I know nothing, but I see a "Troubleshooting"
chapter in the "AdvFS Administration" manual:

http://h30097.www3.hp.com/docs/base_doc/DOCUMENTATION/V51B_HTML/ARH96DTE/TITLE.HTM

(There's also a chapter on "Backing Up and
Restoring Data".)

Or, you could start at:

http://h30097.www3.hp.com/docs/

and fight your way in.

[BrianT_1]

Tried using advscan and that looks good. salvage says "nothing to salvage", yet mount_advfs says that the domain had a problem. Any other suggestions?

[Steven Schweda]

> Any other suggestions?

Recreate the Advfs domain and restore from
backup?

If the disk gets sufficiently corrupted in
sensitive places, then the fix-it tools won't
be able to find enough info to know what to
do.

> Tried using advscan and that looks good.

Define "looks good".

> [...] salvage says "nothing to salvage"
> [...]

Is it happy because the data look good, or
because it can't find any data about which to
complain?

[BrianT_1]

Do any of you know of a consultant we can hire to help with this? We're located in Michigan. I just don't know enough to proceed on my own.

Acually, it's not salvage that says "nothing to fix". Running /sbin/advfs/advscan -f says, literally, "nothing to fix".

[Rob Leadbeater]

Hi,

Recovery of /usr should be fairly straightforward, assuming you have good backups...

Do you have tapes to restore from ?

Cheers,

Rob

[BrianT_1]

While I don't recall configuring backups (it's been a LOOOONG time since I configured this system - I'm not a Unix admin normally, but I can read and follow instructions and that's how I was able to set the system up originally), I can see on another disk several files that are gzipped tar files, one of whose name is fulluser20oct09.tar.gz, and a couple siimilar files with progressively earlier dates, so I think those are the backups. If you can advise me on how to recover the /usr file system, at least I'll be able to check on the contents of those files.

[Rob Leadbeater]

OK. First you need to check that all disks are working as normal. It might be that /usr won't mount because the disk has gone bad.

From the SRM prompt (>>>) can you post the output of "show dev" and "show boot*"

If you then boot to single user, "boot -fl s" can you post the output of "ls -lR /etc/fdmns" and "cat /etc/fstab"

Cheers,

Rob

[BrianT_1]

The disk that housed /usr is the book disk and also houses /. /etc/fdmns was fine. /sbin/advfs/advscan, when used on the domain says "nothing to fix". showfdmn says "E_DOMAIN_NOT_ACTIVATED". I used the File Administration Guide to rebuild the files in /etc/fdmns manually, after using dd to make a copy of the partition that represented /usr (rz8g). /etc/fstab was fine. Nothing in it had changed.

After deleting the domain files and recreating them following advice in the File Administration Guide, I was able to finally mount /usr, but it's completely empty. This is not a good thing. If I restore the contents of the partition with dd in the reverse direction with the file I had created with dd before this started, would that restore the files in /usr?

[Rob Leadbeater]

Hi,

If /usr fails to mount, then I wouldn't expect a dd of it to be any good. You'll have to use your backup. To restore this, you're likely to have to boot up of an OS CD, as (I think) the gunzip utility is usually to be found somewhere under /usr.

So...

1. Boot up off CD, and drop to a Unix shell.
2. Create the relevant symbolic links under /etc/fdmns to be able to mount your existing disks.
3. Use something like "gunzip -c fulluser20oct09.tar.gz | tar tvf - | more" to be view the contents of your backup file.
4. Assuming the /usr file system is in the backup, the restore it with something like "gunzip -c fulluser20oct09.tar.gz | tar xvf - /usr"

Cheers,

Rob

[BrianT_1]

I tried booting off the CD and the first messages I received were "preparing to install". I aborted the process. I don't want to reinstall yet. How can I just mount the CD so I can copy the contents of /usr to the hard drive? Should I even consider that?

[Martin Moore]

When the install CD first boots up, there is an option to exit to the Unix shell. In the GUI version of the installer this is on the "File" pull-down menu.

However, if the issue is restoring /usr, why would you boot from CD? If root is OK, you should be able to boot it to single-user and restore /usr from that point. Booting from CD to restore should only be necessary if you're restoring root.

The sysadmin manual has very good directions on restoring filesystems. The 4.0F manual is at http://www.tru64unix.compaq.com/docs/base_doc/DOCUMENTATION/V40F_HTML/APS2RFTE/TITLE.HTM -- see section 11.4, especially 11.4.1 and 11.4.5. If root is OK, you can skip the first several steps in 11.4.5.

Martin

[Rob Leadbeater]

> Booting from CD to restore should only be
> necessary if you're restoring root.

If the backup files are gzipped, then (I think) you'll need to have a working /usr in order to get the gunzip command, hence why I said to boot off CD.

Cheers,

Rob

[Martin Moore]

To Rob: Ah, OK. I missed that detail in the thread and was assuming that the backups were from something like vdump. If the backups are gzipped, than I agree that he'll need to boot off CD to use gunzip.

To the original poster: Don't panic when it says "preparing to install". The installation process gives you plenty of warning before it actually starts writing over disks. Let it boot up, then choose "Exit to shell" from the File menu, then follow the rest of the procedure Rob suggested.

[BrianT_1]

Since, as I said, advscan, when asked to fix the usr_domain domain, says
"Attempting to fix link/dev_count for domain usr_domain. Nothing to fix", and advscan rzg shows "usr_domain. Domain Id 3f6f4c86.0008beb0. Created Mon Sep 22 15:24:54 2003. Domain volumes 1. /etc/fdmns links 1. Actual partitions found rz8g", but mount_advfs -F 16384 usr_domain#usr /usr gives the ""Domain usr_domain not activated - inconsistency detected, usr_domain#usr on /usr: device does not contain a valid ADVFS file system" and showfdmn usr_domain shows "unable to get info for domain usr_domain. error=E_DOMAIN_NOT_AXTIVATED", is there any way to manually resolve the inconsistency by altering some bits somewhere to remove the inconsistency? How can I even find our what the inconsistency is? I think the data in the partition is intact and if there's any way to recover it without having to reconstruct the entire file system, it would be a good thing.

[BrianT_1]

The Alpha is operational again. I wound up needing to reinstall from CD and after a lot of work, managed to restore most of what I needed. When I didn't have was a backup of a few of the device drivers so I couldn't make a new kernel containing the devices, but fortunately, I had a copy of the pre-crash kernel.

Thanks for the help. I do apreciate it.

[BrianT_1]

Reinstalled from CD and recovered most everything else from backups or the Internet.