Red Hat Bugzilla – Bug 34652
NSCD breaks gethostbyname() error behaviour
Last modified: 2016-11-24 09:48:04 EST
When NSCD is running, gethostbyname() calls it, and gets
back response which treats original DNS SERVFAIL status
as if it got NO-DATA (that is, either NXDOMAIN, or NOERROR without
data) by yielding HOST_NOT_FOUND.
It *should* yield TRY_AGAIN for such an entry.
When I kill nscd processes, the gethostbyname() yields TRY_AGAIN
as a result for the DNS lookup, as it should.
At the time of this reporting, the domain used at the testing was
ICE.QAX.net, but if those folks fix their DNS, things will be different.
This is *probably* generic problem at all Linux versions.
To my amazement, I have seen even Sun Solaris to behave the same!
(The component was wrong, it should have been glibc, not nscd. This
is why this bug went unnoticed.)
I don't think this is a problem for a long time. If you still find
problems, open a new bug. This one is outdated and I'll close it.