Heyo,
I performed sysupgrade on two locally-available OBSD systems on
Tuesday, which went as smoothly as expected, and then initiated
sysupgrade on a machine in colocation. It came back up (per ping), but
refused ssh connections to allow me to complete the upgrade (pkg_add
-u). Investigation (port scanning via nc -nv) showed that the machine
was up and running at least the imap and dns daemons (on both live
interfaces), but not sshd (on either). Daily email (mine are adjusted
to restore disk usage information when that was removed a few releases
back) shows that I _read but failed to heed warnings about the size of
/usr_, so the filesystem is now too full:
OpenBSD 7.6 (GENERIC) #332: Mon Sep 30 08:45:17 MDT 2024
deraadt@amd64.openbsd.org:/usr/src/sys/arch/amd64/compile/GENERIC
1:30AM up 6:52, 0 users, load averages: 0.23, 0.05, 0.02
Running daily.local:
disks:
Filesystem Size Used Avail Capacity iused ifree %iused
Mounted on
/dev/sd1a 1005M 143M 812M 15% 2160 153742 2% /
/dev/sd1h 149G 2.8G 139G 2% 1970 9868364 1%
/home
/dev/sd1p 181G 1.8G 170G 2% 72 11959222 1%
/pub
/dev/sd1d 3.9G 10.0K 3.7G 1% 6 545656 1% /tmp
/dev/sd1f 2.0G 2.0G -87.0M 105% 18324 267498 7% /usr
/dev/sd1g 1005M 464M 491M 49% 9628 146274 7%
/usr/X11R6
/dev/sd1l 19.7G 587M 18.1G 4% 13812 2636554 1%
/usr/local
/dev/sd1o 3.9G 2.0K 3.7G 1% 1 545661 1%
/usr/obj
/dev/sd1j 2.0G 527M 1.4G 28% 191282 94540 67%
/usr/ports
/dev/sd1i 2.0G 1.1G 743M 62% 118844 166978 42%
/usr/src
/dev/sd1m 2.0G 653M 1.2G 35% 32969 252853 12%
/usr/xenocara
/dev/sd1n 3.9G 2.0K 3.7G 1% 1 545661 1%
/usr/xobj
/dev/sd1e 58.6G 21.0M 55.7G 1% 1169 7845997 1% /var
(this one is part of the email from overnight Tue-Wed, which also has
this at the bottom:
Services that should be running but aren't:
sshd
)
So. I have found the problem, yes? And the solution is ... what? Note:
there's a short version at the bottom of the email in the last para; I
prolly talk too much.
Presumably I need console access, since sshd isn't running and I don't
have an exploit to get a shell by breaking dovecot or nsd. But what
then? Slice k is available on /dev/sd1, and either there's free space
(it's prolly 500GB, and slice b is prolly 32GB), or if not then the
contents of slice p (/pub) can be moved to slice h (/home) and the
space now allocated to p can be reduced by circa 6G to allow creation
of a /usr that's more appropriately sized.
But will that work? Is the overfull (-87M) /usr complete? Will copying
it over to a new location create new, intermittent, can't-be-reproduced
bugs?
Is a complete/clean reinstall mandated? Nightly backups to sd2a are
working, so there's configuration information (and a presumably
bootable kernel) in multiple locations; that config information (on an
unmounted drive) should survive a reinstall and provide configs for
rebuilding. Since /altroot has the same 7.6 kernel, is it possible to
bring up root or altroot single user, with at most /tmp mounted, and
tinker with partitions that way? Or does the overfull result indicate
that a clean install is required?
Is there a way to get access remotely? I'm pretty sure that there
isn't, because duh. I presume that it's not coming up single user
(because other daemons appear to be running, including at least crond,
dovecot, nsd, and (probably) unbound), and all the filesystems
(including other disks not shown above) are up. My colo guy isn't
responding to texts or emails currently, but that's a social, not a
technical problem. It may even be lucky, since I didn't notice the
overfull problem until my attention was pointed to it this morning by
discussion in the other thread, so I haven't tried hacking on things
without an understanding of the problem.
I'm too verbose. Short version: a) with sshd not working and no console
access, is there a way to work on it, or is console access required? b)
assuming (prolly console) access, can the current contents of /usr be
copied to a new /usr slice, and the rest of the upgrade (pkg_add -u)
performed, or is a complete/clean install required for stability?
Thanks for your time,
Amy!
--
Amelia A. Lewis amyzing {at} talsever.com
Confidence: a feeling peculiar to the stage just before full
comprehension of the problem.
No comments:
Post a Comment