[Openais] Missing shutdown log entry corosync 1.2.1 (Reminder)

Andreas Mock Andreas.Mock at web.de
Wed Apr 28 01:56:42 PDT 2010


-----Ursprüngliche Nachricht-----
Von: Jan Friesse <jfriesse at redhat.com>
Gesendet: 28.04.2010 09:33:39
An: Andreas Mock <Andreas.Mock at web.de>
Betreff: Re: [Openais] Missing shutdown log entry corosync 1.2.1 (Reminder)

>Andreas,

Hi Jan,

>from strace it looks like you are using new corosync exec with old
>libcoroipcs.so
>
>Can you please:
>nm /usr/lib64/libcoroipcs.so.4.0.0  | grep coroipcs_ipc_service_exit
>
>(or /usr/lib - depends on system you are using)
>
>output should be:
>
>00000000000030c5 T coroipcs_ipc_service_exit

It's there but with different offset.

As soon as I read your hint the scales fell from my eyes.
Last year there was no current version available for openSuSE 11.2 and
I started compiling corosync by myself. As soon as Andrew promised
to package newer versions for openSuSE 11.2 I wanted to avoid
the hassle of doing it all myself and changed to the rpm packages.
As long as the version were the same I didn't recognized that
the rpm binaries loaded the self compiled shared libraries from
a totally different directory (/usr/local/lib comes  before /usr/lib64 in
/etc/ld.conf, don't ask me why)

So fact is: After cleaning the old mess and doing a ldconfig -v on both
nodes I do get the messages as expected.

So I have to say a big THANK YOU for investigating this problem and
spending your time for a problem caused by a stupid administrator.

Based on this case a question to avoid these kind of strange errors:
Would it be feasible to check the versions of all loaded shared libraries
or services at startup/runtime. A kind of (plugin) API version conformance test?

Once again: Thank you

Best regards
Andreas Mock


More information about the Openais mailing list