Description of problem:
In 6.2, setting entry_cache_nowait_percentage would allow SSSD to update
often-used entries out-of-band, so the cache would be refreshed outside of
requests and users would experience fewer cache-misses. In 6.3, this feature no
longer works.

Version-Release number of selected component (if applicable):
sssd-1.8.0-27.el6

How reproducible:
Every time


Steps to Reproduce:
1. Set entry_cache_nowait_percentage = 50 in the [nss] section of sssd.conf
2. Set entry_cache_timeout = 20 in the [domain/DOMAINNAME] section of sssd.conf
3. Create a valid netgroup entry in LDAP (I'll use 'netgroupname' below.
4. Purge your cache and restart sssd (to ensure clean starting state)
5. Run the following bash command:
for i in {1..20}; do time getent netgroup netgroupname; sleep 6; done

Actual results:
The first request takes a little while (depending on network conditions) to
return. The second and third will return virtually instantly, the fourth  will
take longer again, and this pattern will repeat.

Expected results:
The first request takes a little while (depending on network conditions) to
return. All subsequent requests (as long as they keep being requested in 6s
periods) should return virtually instantly.

Additional info:
Technical reason: We added an in-memory cache for netgroups to reduce the
amount of time we needed to spend processing results from the sysdb cache.
However, we set the lifetime of this in-memory cache to the full lifetime of
the sysdb cache entry, instead of taking the nowait percentage into account.
There is a patch under review upstream now that will address this issue.

Setting priority to high and Keyword Regression because if this is released in
6.3 final, it WILL necessitate an errata to correct it, as many customers
depend on this feature for performance reasons.