Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Thu, Apr 5, 2012 at 3:30 PM, Lars Ellenberg <lars.ellenberg at linbit.com> wrote: > On Thu, Apr 05, 2012 at 08:56:52AM +0000, Maurits van de Lande wrote: >> >Well you've mentioned here you're already working with Linbit, so >> >what did they say about this? Surely the same issue would appear >on >> >vmhost6a and vmhost6b whose configurations you've previously posted >> >here >> Yes the same issue occurs on all vmhostxx clusters. Linbit was looking >> at the performance issues, they didn't check that bug report. Maybe I >> should ask Linbit again. When I posted the bug at elrepo they said >> they think it is a harmless warning. >> >> Hopefully we get a reply from Linbit regarding this subject. > > It is a harmless warning. > The impact is neither barriers nor flush/fua will be available. > > It is caused by RHEL kernel being "kABI compatible" in theory, > but in fact changed the semantics of bio bitfields completely, > between 6.0 and 6.1. OK. Then the situation is exactly the same as I explained here months ago: http://lists.linbit.com/pipermail/drbd-user/2011-August/016628.html And funnily enough, that same email was also quoted here: http://elrepo.org/bugs/view.php?id=210 > Packages built by Linbit do not trigger this warning, > and work correctly, afaik. > > I can't say what the centos packages do wrong, > but I assume they have been built against 6.0 kernel headers, > and now loaded into a 6.1 or later binary kernel. > > I suggest to rebuild against the correct headers. That, too, is in http://elrepo.org/bugs/view.php?id=210 So it seems we need a cluebat for the ELrepo folks. Or is this something that's really easy to get wrong during a build? Florian