It's not running in production, and @BBlack says we should rip it out unless absolutely needed.
Investigate why it was introduced and see if it still is required.
It's not running in production, and @BBlack says we should rip it out unless absolutely needed.
Investigate why it was introduced and see if it still is required.
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | valhallasw | T100554 Grid engine masters down | |||
Declined | yuvipanda | T100564 Investigate why nscd is used in labs |
It was introduced with de059228933681b6b0f97a818a561cde22901e1e ("Initial commit of public puppet repo."). In the past IIRC we have increased the caching TTLs several times to reduce network load for DNS requests. With the changes in the network configuration (conntrack), it may be possible for the Labs network to sustain the load now, but that issue should be approached with due caution :-).
I'd say mostly to avoid roundtrips to ldap?
$ sudo nscd -g | grep -e 'cache:$' -e 'rate' -e 'number of cached' passwd cache: 1% cache hit rate 38 current number of cached values 948 maximum number of cached values group cache: 0% cache hit rate 54 current number of cached values 83 maximum number of cached values hosts cache: 0% cache hit rate 14 current number of cached values 23 maximum number of cached values services cache: 0% cache hit rate 1 current number of cached values 6 maximum number of cached values netgroup cache: 0% cache hit rate 0 current number of cached values 0 maximum number of cached values
It's back on on tools-master now, and doesn't really seem to be causing any issues atm.
nslcd AFAIK doesn't cache responses, so querying nss will still result in a ldap lookup (through nslcd) if nscd isn't running