http://bugzilla.novell.com/show_bug.cgi?id=1121221 Bug ID: 1121221 Summary: NFSv4 Kerberos Subtree Check performance problem Classification: openSUSE Product: openSUSE Distribution Version: Leap 15.0 Hardware: Other OS: Other Status: NEW Severity: Normal Priority: P5 - None Component: Basesystem Assignee: bnc-team-screening@forge.provo.novell.com Reporter: ralf.koelmel@kit.edu QA Contact: qa-bugs@suse.de Found By: --- Blocker: --- I've tested in my mixed setup NFSv3, NFSv4 with sec=sys or sec=krb5 (with openSUSE Leap 15.0 on both sides) and found only with NFSv4 and Kerberos in all 3 variants (nfsvers=4.0,4.1,4.2) the following problems: 1. with nfsvers=4.0,sec=krb5: mounting fast, but bad performance (factor 3 worser as with sec=sys) with randomly occuring huge delays (>15 seconds) during directory traversal (with "time find . >/dev/null" in a home directory and a lot of subdirectories) 2. with nfsvers=4.1 or 4.2,sec=krb5: very slow mounting (nearly 2 minutes); after the mountpoint is available, the performance is good and there is no delay during directory traversal Also changing from rpc-svcgssd to gssproxy doesn't change anything. In tcpdump traces on the serverside during the 2. testcase i've seen the following errors every 10 seconds until the mountpoint is available after almost 2 minutes : " NFS reply xid ... reply ok 140 getattr ERROR: Request couldn't be completed in time " With NFSv3 sec=krb5 or sec=sys or NFSv4 with sec=sys i've found no problems. I've found that the NFS export default option "subtree_check" is influencing these problems. With setting the option "no_subtree_check" on the NFS export definitions the problems with NFSv4 and Kerberos are gone. That Subtree Checking can result in worser performance is clear, but it should not be so bad. In my opinion this is a bug in combination with NFSv4 and Kerberos. -- You are receiving this mail because: You are on the CC list for the bug.