From: Xiaoguang W. <wan...@cn...> - 2014-05-13 11:58:12
|
Hi, Recently I met some fails in RHEL7.0Beta by running "./runltp", the fails are below: ------------------------------------------ hugemmap01 FAIL 2 hugemmap03 FAIL 2 hugemmap04 FAIL 2 ------------------------------------------ After some investigation, I found that it is "echo 128 > /proc/sys/vm/nr_hugepages" causing these fails. By catting /proc/buddyinfo or catting /proc/pagetypeinfo, I found that the system has too many memory fragments and was not able to allocate 128 hugepages. So I also want to know which test cases results in these too many memory fragments. It seems that LTP does not provide a method to get the system's status information between these tests. I think we may add patches to support providing a shell script to ltp-pan, which will be called before and after every test case, to record some system status information. In this shell script, you can collect system information you has interests in, then save these information to files. Then we can analysis these files to know which case causing the big system status changes. Anybody has some better ideas, thanks. Regards, Xiaoguang Wang |