On 06/12/2012 01:53 PM, 686f6c6d wrote:
On Tue, Apr 10, 2012 at 1:19 PM, 686f6c6d686f6c6d@googlemail.com wrote:
On Mon, Apr 2, 2012 at 16:16, 686f6c6d686f6c6d@googlemail.com wrote:
- Using kernel-default-3.1.0-1.2.1.x86_64 from the "install" repo, the
installation succeeds.
- Using kernel-default-3.1.9-1.4.1.x86_64 from the "update" repo, the
installation fails.
Let me correct that: Once I changed the profile again to re-include some of our scripts (that were disabled while debugging), even kernel-default 3.1.0 fails with the same problem. Need to do further debugging.
When working around the kernel issue by using only the suse "install" repo (and not the "update" repo), I see the following behaviour:
- /var/log/messages show these errors: "su: pam_systemd(su:session):
Failed to create session: Invalid argument" (seems to be https://bugzilla.novell.com/show_bug.cgi?id=731358 )
- Although "network", "ntp" and "sshd" are shown by chkconfig to be
"on" (with ntp and sshd being explicitly specified as "3 5" in <runlevel>), they are not started at boot ("systemctl list-units -t service --all" shows them all as "loaded inactive"). Do I have to change anything else to make yast2-runlevel work with systemd?
- The network comes up during boot (it works flawlessly in stage2 and
/etc/sysconfig/network/ifcfg-eth0 and [...]/routes look fine), but then seems to exit for no particular reason, even when I try to start it manually (with "systemctl start network.service"). The only thing in the logs is this: ---------- 8< ---------- ifup: eth0 ifup: IP address: [...] dns-resolver: ATTENTION: You have modified /etc/resolv.conf. Leaving it untouched... dns-resolver: You can find my version[...] ifdown: eth0 ---------->8 ---------- systemctl shows it as "Active: inactive (dead)" then, with "ExecStart=[...], status=0/SUCCESS". I am *so* clueless what happens.
I've seen this a lot, but not all the time, when upgrading systems from 11.3 to 12.1, via "zypper dup". From what I can gather the problem seems to be due to "netconfig update" exiting with a non-zero status due to resolv.conf not matching the NETCONFIG_DNS_POLICY. I've been mostly successful in working around this by setting NETCONFIG_DNS_POLICY="" in /etc/sysconfig/network/config. I say mostly successfully because the "inactive (dead) networking has happened even after setting NETCONFIG_DNS_POLICY="".
This seems to happen more on VM's than on physical servers for me. When dup'ing to 12.1 the VM's don't always like to reboot gracefully so I was doing "reboot -f" after which networking may or may not be "inactive (dead)". I found that when it was dead if I stopped networking and rebooted the VM, networking would function properly subsequently. Also, after dup to 12.1 if I don't "reboot -f" and simply reboot, and once the system hangs up during shutdown if I reset power to the VM networking comes up fine.
Since there's really no verbose|debug logging|output in systemd/systemctl it's REALLY difficult to get to the root of the problem. If anyone has any ideas I'm all ears? I haven't gotten to strace'ing the network.service yet but I have a feeling it may be only way to get an really info about what going on.
(When I forbid systemd-sysvinit in autoyast, everything works fine as expected again.) I really hope 12.2 will be delayed, otherwise my resume will read "failed to make autoyast work with systemd for 12.1's entire lifetime." (;