[Bug 1378691] [NEW] nfs4 id mapping unstable

[Andrew]

Public bug reported:

Ubuntu 14.04 LTS
Mounting an NFS4 share from a EMC nas and using ldap to get ids through nsswitch.

This works but occasionally an id is lost suddenly and ls -l shows the
file owner and group as nfsnobody.

I tried using sysctl to increase kernel.keys (it didn't help)
kernel.keys.root_maxkeys = 10000
kernel.keys.root_maxbytes = 200000
kernel.keys.maxkeys = 10000
kernel.keys.maxbytes = 200000

I tried changing the timeout in /etc/request-key.d/id_resolver.conf to 10 seconds (it didn't help)
create  id_resolver     *       *       /usr/sbin/nfsidmap -t 10 %k %d

To fix the problem I have manually upgraded the kernel to
3.13.0-36-generic and I continually clear the id keys using "watch
nfsidmap -c".  This seems to have reduced the problem so that I haven't
seen it so far after 2 hours.

It seems that occasionally a bad id lookup happens, and then gets cached
as bad for a very long time, and that my clearing the cache manually
gets it to do another lookup. (Does the key return bad when it times-out
instead of re-looking up?)

root# cat /proc/keys
...
29dccc9d I--Q---     1   9m 3b010000     0     0 id_legacy uid:aj-hart at uwe.ac.uk: 5

and that the "-t 10" setting is ignored and is set to 600~9m instead.

I am hoping for a better fix than my "watch nfsidmap -c" hack if
possible.

** Affects: nfs-utils (Ubuntu)
     Importance: Undecided
         Status: New

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to nfs-utils in Ubuntu.
https://bugs.launchpad.net/bugs/1378691

Title:
  nfs4 id mapping unstable

Status in “nfs-utils” package in Ubuntu:
  New

Bug description:
  Ubuntu 14.04 LTS
  Mounting an NFS4 share from a EMC nas and using ldap to get ids through nsswitch.

  This works but occasionally an id is lost suddenly and ls -l shows the
  file owner and group as nfsnobody.

  I tried using sysctl to increase kernel.keys (it didn't help)
  kernel.keys.root_maxkeys = 10000
  kernel.keys.root_maxbytes = 200000
  kernel.keys.maxkeys = 10000
  kernel.keys.maxbytes = 200000

  I tried changing the timeout in /etc/request-key.d/id_resolver.conf to 10 seconds (it didn't help)
  create  id_resolver     *       *       /usr/sbin/nfsidmap -t 10 %k %d

  To fix the problem I have manually upgraded the kernel to
  3.13.0-36-generic and I continually clear the id keys using "watch
  nfsidmap -c".  This seems to have reduced the problem so that I
  haven't seen it so far after 2 hours.

  It seems that occasionally a bad id lookup happens, and then gets
  cached as bad for a very long time, and that my clearing the cache
  manually gets it to do another lookup. (Does the key return bad when
  it times-out instead of re-looking up?)

  root# cat /proc/keys
  ...
  29dccc9d I--Q---     1   9m 3b010000     0     0 id_legacy uid:aj-hart at uwe.ac.uk: 5

  and that the "-t 10" setting is ignored and is set to 600~9m instead.

  I am hoping for a better fix than my "watch nfsidmap -c" hack if
  possible.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/nfs-utils/+bug/1378691/+subscriptions