Alf Wachsmann wrote:
I am trying to monitor one of our machines where we see these kind of kernel panics with "afsmonitor". On the machine itself "kdump -users" shows right now around 690 'afs_users' entries. When I use "afsmonitor" and look at the values of "curr_PAGs", "PAGCreations", etc., they are all 0. Is there a bug in how this information is made available to "afsmonitor"?
Lol! Depends: yesterday we found that on x86_64 xstat_cm_test no longer works since the kernel stores a struct timeval as two afs_int32 (osi_timeval_t) whereas user space still believes the system's 'struct timeval' is the right thing. I'm currently patching this.
The reason I mention it is that afsmonitor uses the same interface into the kernel and not unlikely royally ignores which length the returned array of numbers is, which could explain unexpected values. xstat_cm_test is a bit more careful and refuses to cooperate.
On i386 there should be no problem. Solaris and other "real" Unixes I don't know, things like Irix may actually need different user space binaries depending on whether they run 32bit or 64bit kernels (arrrgh).
-- =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= Rainer Toebbicke European Laboratory for Particle Physics(CERN) - Geneva, Switzerland Phone: +41 22 767 8985 Fax: +41 22 767 7155 _______________________________________________ OpenAFS-devel mailing list OpenAFS-devel@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-devel
