You are not logged in.
Hi all.
I am dealing with a similar situation as of today after upgrading, only this is a headless server.
The nss update was according to arch-news instructions.
Removed quiet boot, all seems ok with no errors but after loading login service, I get a blinking cursor.
This is a headless server with a nextcloud lemp instance. I had to move keyboard and screen to it in order to boot with -b option
But not much help.
Restored from image backup so I am back to stage one and can ssh to the server.
The updates were the following
sudo pacman -Syu
:: Synchronizing package databases...
core 135.0 KiB 1350 KiB/s 00:00 [######################] 100%
extra 1648.2 KiB 1249 KiB/s 00:01 [######################] 100%
community 4.9 MiB 1455 KiB/s 00:03 [######################] 100%
multilib 161.0 KiB 1610 KiB/s 00:00 [######################] 100%
:: Starting full system upgrade...
warning: php: ignoring package upgrade (7.3.12-1 => 7.4.5-1)
warning: php-apcu: ignoring package upgrade (5.1.18-1 => 5.1.18-2)
warning: php-fpm: ignoring package upgrade (7.3.12-1 => 7.4.5-1)
warning: php-gd: ignoring package upgrade (7.3.12-1 => 7.4.5-1)
warning: php-igbinary: ignoring package upgrade (3.0.1-2 => 3.1.2-1)
warning: php-imagick: ignoring package upgrade (3.4.4-4 => 3.4.4-6)
warning: php-intl: ignoring package upgrade (7.3.12-1 => 7.4.5-1)
warning: php-redis: ignoring package upgrade (5.1.1-2 => 5.2.1-1)
resolving dependencies...
looking for conflicting packages...
warning: dependency cycle detected:
warning: systemd will be installed before its ca-certificates-mozilla dependency
Packages (59) automake-1.16.2-1 bison-3.5.4-1
ca-certificates-mozilla-3.51.1-1 coreutils-8.32-1
device-mapper-2.02.187-1 dhcpcd-9.0.1-2 dialog-1:1.3_20200327-1
e2fsprogs-1.45.6-1 gd-2.3.0-1 git-2.26.2-1 glib2-2.64.2-1
gnupg-2.2.20-1 gnutls-3.6.13-1 go-2:1.14.2-1
graphite-1:1.3.14-1 guile-2.2.6-2 harfbuzz-2.6.5-1
hwids-20200306-1 hwinfo-21.69-1 iana-etc-20200401-1
imagemagick-7.0.10.7-1 iproute2-5.6.0-1 krb5-1.18-1
libffi-3.3-3 libnetfilter_conntrack-1.0.8-1 libnftnl-1.1.6-1
libnghttp2-1.40.0-1 libp11-kit-0.23.20-4 libsecret-0.20.3-1
libtirpc-1.2.6-1 libutil-linux-2.35.1-2 libzip-1.6.1-1
linux-firmware-20200316.8eb0b28-1 linux-lts-5.4.33-3
linux-lts-headers-5.4.33-3 make-4.3-3 nano-4.9.2-1
netctl-1.21-2 nginx-mainline-1.17.10-1 nss-3.51.1-1
openssl-1.1.1.f-1 p11-kit-0.23.20-4
pacman-mirrorlist-20200411-1 perl-5.30.2-1 python-3.8.2-2
python-cffi-1.14.0-2 python-configargparse-1.2.1-1
python-pyparsing-2.4.7-1 python-setuptools-1:46.1.3-1
python-urllib3-1.25.9-1 python-zope-component-4.6.1-1
python-zope-interface-5.0.2-1 redis-5.0.9-1
reflector-2020.3.21.2-1 systemd-245.5-1 systemd-libs-245.5-1
systemd-sysvcompat-245.5-1 util-linux-2.35.1-2 xz-5.2.5-1
Total Download Size: 413.15 MiB
Total Installed Size: 1572.44 MiB
Net Upgrade Size: 3.41 MiB
:: Proceed with installation? [Y/n] nphp updates are ignored because the upgrade breaks nextcloud (help would be welcome)
Please advise what info may be needed at this time to address the issue, observing that the server is currently at previous state-backup (before update is applied)
Thanks in advance.
Last edited by justm3 (2020-04-21 13:19:09)
Offline
a) please use code tags, not qzote tags
b) this thread revolved around a failed update thanks to a … questionable update service. Afaiu that's not the case for you, so it's unlikely related.
Open a new thread presenting your case rather than hijacking this one (once the OP marks it solved, nobody else will really look at it)
There've been several issues around the dhcpcd-9.x update, so if you're currently usign an 8.x version, try whether ignoring that package still breaks the update. If it's dhcpcd, there're updated versions (9.0.1-4) in the testing repo that are claimed to fix various issues, you could download the package and install it for a test w/ "pacman -U"
nb. that those are just explorative measures and that partial updates are generally not supported.
Offline
Noted, corrected code tags. My bad.
dhcpcd is 8.1.6
I can exclude it. The thing is that server restore is a time-pain.
What is your opinion?
Offline
If you're sure you didn't botch the p11-kit-trust.so update, dhcpcd is the most likely contender.
You've two kernels so even if one of them was the culprit, you'd simply (?) boot the other one.
General reasons for failing boots are https://wiki.archlinux.org/index.php/Microcode and https://wiki.debian.org/BoottimeEntropyStarvation (also see https://wiki.archlinux.org/index.php/Rng-tools ) but that would be quite a coincidence
Offline
@seth you stand correct.
I took a deep breath, excluded dhcpcd and update was successful.
Thank you so much, you are a star.
P.S. I know it is not my field but should this instance be mentioned? Maybe in the arch news for others to avoid it? Or maybe it is done and I missed it.
Offline
There's https://bugs.archlinux.org/task/66321 but it's not clear what cases are covered by the patched version.
I doubt there'll ever be a news item that says "don't update package X to the current version", the maintainer in charge could rather revert or update the broken package.
I'd ask you try 9.0.1-4, but there's a good chance the system will still break, so if accessing it is a complete PITA, rather don't. Wait and hope for the best.
Upstream is most likely very awayre of the situation, because there're recent related commits and he's also on the forum ;-)
Offline
Well I' rather not try it. The server is in production, I would not obstruct users any further.
When the machine was back in 4.19-lts, I was getting some messages from dhcpcd. I am not sure but I think it was the maintainer who told me not to bother or upgrade to a newer kernel. Which I did once 5.4-lts was released and journalctl cleared.
The thing is I cannot switch over to networkmanager because nginx complains.
So be it dhcp. It works just fine.
Again thank you so much.
Be well.
Offline