on yet another retry the failures in sanity did not reproduce, but some of the fails in sanity-sec did.
https://testing.hpdd.intel.com/test_sessions/cc30db56-3c30-11e5-af2f-5254006e85c2
Bob, is this failure always related to the nodemap feature,or only in the sanity-sec tests?
Also, is the problem related to the 6.7 client or the 6.7 server?
Andreas., the 'Error: 'adding fops nodemaps failed 1' is seen on the sanity-sec fails, not on other fails. Not clear to me if they are due to client or servers. all 6.7 runs reported have 6.7 on both client and server.
sometimes the error text is Error: 'nodemap_add failed with 1'
always in sanity-sec
Did something change with proc fs? They all seem to fail after lctl get_param.
Most tests failed relate to proc fs. There haven't a significant difference between 6.6 & 6.7 fs codes. But i suspect some scheduler changes maybe impact proc fs behaviors. For example, the sanity-sec test_7 failure looks like read a proc entry after create. maybe some create action still not finished while read coming. This test case just operating on MGS node locally, so haven't any network & lock been involved. I have submit a debug patch(http://review.whamcloud.com/15916), hope hit a instance.
there's now at least 1 counterexample. Here is a test run with el6.7 on master with no fails in either sanity or sanity-sec;
https://testing.hpdd.intel.com/test_sessions/e835d3fe-4074-11e5-8abf-5254006e85c2
As far as I know there were no significant changes from all those repeated fails and this one good run.
bogl: And now there is this errata kernel. Maybe you will get lucky and it will resolve your issues here. 
Brian: thanks, I'm on it. see latest comment in LU-6894. The new kernel arrived in Centos late yesterday. revised 6.7 mods in flight.