From Bugzilla Helper: User-Agent: Mozilla/4.77 [en] (X11; U; Linux 2.4.8 i686) Description of problem: nscd is started with 6 threads by default (nscd.conf). while debugging some other problems i started it through strace with -t 1 -d and observed immediate lockup when a uid/gid should get resolved from ldap. Version-Release number of selected component (if applicable): redhat 7.1.94 (glibc and nscd 2.2.4-5) How reproducible: Always Steps to Reproduce: 1. have a ldap based uid/gid resolving and some nfs dir mounted somewhere, with files/dirs owned by uids from ldap. 2. start nscd with one thread only 3. do a ls -l in the nfs mounted dir Actual Results: ls -l locks forever (at least until nscd is stopped) while ls -ln works. nscd actualy says that the first uid the ls -l should resolve is not in it's cache, but that's as far it gets. Expected Results: i would expect the decrease in number of threads would only impact peformance, not functionallity. but there's something odd about it, it works ok if i say -t 2. could this be related to other nscd lockups people are expiriencing (including me)? Additional info: mostly the same as at #54646
This problem should now be fixed in the official glibc CVS sources. The next binary we produce should contain the fix.
I'm seeing this too with NIS+ with the default number of threads. Will there be a new binary for RH9? Or are the glibc+nscd RPMs in rawhide stable enough to use in production? Thanks.
Try the binaries at ftp://people.redhat.com/jakub/glibc/errata/2.3.2-27.9.4/ which are supposed to be the next RHL9 errata.