High CPU usage in sssd_autofs running Chef knife

[John Beranek]

So, recently noticed an issue, which I think is new.

Fedora 32
sssd 2.3.0 - configured for AD, including sudo and autofs

I found that when Chef's "knife" command (from Chef Workstation 0.3.2) is
run, the command takes a very long time to run, and in top sssd_be and
sssd_autofs use a lot of CPU.

e.g.

    PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+
COMMAND
  17579 root      20   0  259196  32220   9340 S  28.2   0.4   0:18.62
sssd_autofs
  17575 root      20   0  346320  25276  21672 S  24.3   0.3   0:16.54
sssd_be

Turned debug up to 9 for the autofs module and what looked apparent to me
was the following, from one run of 'knife':

]# grep "not found in cache" sssd_autofs.log |wc -l
53144
# grep "not found in cache" sssd_autofs.log |head -n 20
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15184: Object [auto.home] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15184: Object [auto.home] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15185: Object [auto.home:gem.deps.rb] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15185: Object [auto.home:gem.deps.rb] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15185: Object [auto.home:gem.deps.rb] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15185: Object [auto.home:gem.deps.rb] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15186: Object [auto.home:/] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15186: Object [auto.home:/] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15186: Object [auto.home:/] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15186: Object [auto.home:/] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15187: Object [auto.home:*] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15187: Object [auto.home:*] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15187: Object [auto.home:*] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15187: Object [auto.home:*] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15188: Object [auto.home] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15188: Object [auto.home] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15189: Object [auto.home:gems.rb] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15189: Object [auto.home:gems.rb] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15189: Object [auto.home:gems.rb] was not found in cache
(2020-05-27 16:19:24): [autofs] [cache_req_search_cache] (0x0400): CR
#15189: Object [auto.home:gems.rb] was not found in cache

Any ideas what could be causing this? Happy to provide sanitised config if
required.

Cheers,

John

-- 
John Beranek                         To generalise is to be an idiot.
http://redux.org.uk/                                 -- William Blake

[John Beranek]

On Wed, 27 May 2020 at 16:28, John Beranek wrote:
>  So, recently noticed an issue, which I think is new.
>
>  Fedora 32
>  sssd 2.3.0 - configured for AD, including sudo and autofs
>
>  I found that when Chef's "knife" command (from Chef Workstation 0.3.2) is
> run, the command takes a very long time to run, and in top sssd_be and sssd_autofs use a
> lot of CPU. 
Probably pertinent that I'm running knife as a user with an
auto.home-mounted NFS home directory, though I generally don't run
knife from within said NFS home.

With nsswitch.conf having "automount: sss files":

$ time knife node list > /dev/null
knife node list > /dev/null  5.37s user 2.11s system 12% cpu 59.150 total

and with "automount: files":

$ time knife node list > /dev/null
knife node list > /dev/null  4.28s user 1.80s system 82% cpu 7.393 total

Cheers,

John

-- 
John Beranek                         To generalise is to be an idiot.
http://redux.org.uk/                                 -- William Blake