Re: corrupt /etc: how to gracefully shut down?

[Brad Spencer <brad%anduin.eldar.org@localhost>]

hhaamu%gmail.com@localhost writes:

> On Fri, 06 Nov 2026 12:04:05 -0500
> Brad Spencer <brad%anduin.eldar.org@localhost> wrote:
>
>> What I am going to suggest assumes you have console access.  If you do
>> not, then it may be more challenging.
>> 
>> On the console as root:
>> 
>> kill -15 1    - this will bring you down to single user
>> umount -a     - this will unmount as many of the filesystems as it can.
>>               In particular the root filesystem won't be unmounted with this.
>> poweroff      - or equivalent...  shuts the system down
>
> Thanks, I did indeed have console access so I followed this, with the
> change that I sent a SIGTERM to nfsd and its sibling processes and
> unmounted the data partitions (and disabled swap) before sending the
> SIGTERM to init.

None of that would hurt, but I also don't think any of it would really
matter.  Once you are in single user, most processes have been dealt
with unless they are hung up.  NFS is stateless and doesn't hold open
files, I don't think.  I believe, for example, that you can unmount a
exported file system out from under nfsd and friends and if you can do
that then it doesn't matter that nfsd is running.  Swap doesn't really
need to be disabled, especially if you are powering off.  I never
disable it anyway.

> I was under the impression that you had to disable the raid with raidctl
> but could not find a trace of that in the missing rc scripts so I let
> it be and let poweroff handle the rest. The system made a 'thunk' sound
> when it powered off which was unnerving -- I expected to hear the hdds
> spin down a few seconds before the actual power cut. Maybe I should've
> used 'halt' instead and cut power manually. But I guess it worked out
> all right in the end.

I have had to shut down systems in bad ways over the years, and have
never known raidframe to care very much.  Really, poweroff is just "halt
-p" anyway... if you wanted to do a reboot, which seems unlikley in this
case, then powering off probably isn't the thing to do, of course.  Any
spinning rust drive made in the last 30 years (at least) is likely to
auto retract in any case.  I don't know what the thunk might have been
but maybe not the drives.


But in any case, glad that progress was made.  Hope you can fix the
corruption problem.  As someone else pointed out, you might have local
backups of /etc if your crons have been running.  Of course, some of
your corrupted directories could be pulled from the install sets.



-- 
Brad Spencer - brad%anduin.eldar.org@localhost