Alright, I got a little further, apparently I had to uninstall
libcrmcommon as well. So now I can start the newer pacemaker version
on one node, it also connects successfully with the other node. One
resource could be started (galera), but for the rest of them (mainly
systemd units) I get "Not installed" errors like these:
cinder-volume probe on controller01 could not be executed (Not
installed) because 'cinder-volume: No such file or directory'
I'm not sure if this is due to missing crmsh on the node. But if I
installed crmsh via package manager, it would install the older
pacemaker version again. I couldn't find a configure flag for crmsh,
so if you have another pointer, I'd greatly appreciate that!
Zitat von Chris Lumens <clum...@redhat.com>:
I'm clueless, tbh. I removed all pacemaker packages from that host
before running clean and install. If you have another hint for me,
I'd be greatful!
Both of those symbols are in libcrmcommon. Make sure you have that
installed somewhere - ls -l /usr/lib*/libcrmcommon*.
You can also use ldd on the installed programs to verify that they know
how to find the libraries they need. For instance:
$ ldd /usr/sbin/crm_mon
linux-vdso.so.1 (0x00007f6693c3d000)
libpacemaker.so.1 => /lib64/libpacemaker.so.1 (0x00007f6693b94000)
libpe_status.so.28 => /lib64/libpe_status.so.28 (0x00007f6693b31000)
...
If it doesn't show an actual file, you can run ldconfig as root to
regenerate the dynamic linker cache. Sometimes this is necessary when
you install new libraries, and often times the packaging system will do
this kind of thing for you.
- Chris
_______________________________________________
Manage your subscription:
https://lists.clusterlabs.org/mailman/listinfo/users
ClusterLabs home: https://www.clusterlabs.org/
_______________________________________________
Manage your subscription:
https://lists.clusterlabs.org/mailman/listinfo/users
ClusterLabs home: https://www.clusterlabs.org/
