Bugzilla – Bug 426679
nscd (GNU libc) 2.8 - random crash
Last modified: 2008-11-19 19:25:48 UTC
Hi, I'm using openSUSE 11 and since my last kernel update nscd became really unstable. I can't reproduce the error in a easy way, but from time to time nscd stops working and then some applications like firefox and kmail stop working too (mostly because they don't find the DNS server). Some time ago while in Mexico in a client i had a similar problem with nscd in SLES10 (64bits) and it wasn't resolved but it happened just when I had integration with ldap. right now it's affecting just my desktop as i didn't do the upgrade in production machines yet. the /var/log/messages output is: nscd: 19165 invalid persistent database file "/var/run/nscd/passwd": verification failed i already tryed to erase the files in /var/run/nscd/* so nscd could reacreate the database. thank you a lot
Hello all, I've some more info about this bug, logs with debugging: 14499: considering GETHOSTBYNAME entry "www.google-analytics.com", timeout 1221668513 14499: handle_request: request received (Version = 2) from PID 27351 14499: GETFDPW 14499: provide access to FD 4, for passwd 14499: handle_request: request received (Version = 2) from PID 27351 14499: GETFDPW 14499: provide access to FD 4, for passwd 14499: handle_request: request received (Version = 2) from PID 27357 14499: GETFDPW 14499: provide access to FD 4, for passwd 14499: handle_request: request received (Version = 2) from PID 27357 14499: GETPWBYUID (105) 14499: Haven't found "105" in password cache! 14499: add new entry "105" of type GETPWBYUID for passwd to cache (first) 14499: add new entry "haldaemon" of type GETPWBYNAME for passwd to cache 14499: handle_request: request received (Version = 2) from PID 27357 14499: GETPWBYUID (1) Bus error #########################
Please include your /etc/nscd.conf and /etc/nsswitch.conf. Are you *sure* other applications stop working after nscd crash? Normally, this should never happen, only the caching should stop working but the applications are supposed to just keep resolving on their own. Can you set ulimit -c unlimited and then attach a core? Do you get the same kind of crash (like you pasted from nscd -d output) every time it crashes? nscd indeed has stability problems in opensuse 11.0, but normally the crashes look quite differently.
Created attachment 241987 [details] config file for nscd
Created attachment 241988 [details] nsswitch.conf
I did try the ulimit -c unlimited but it doesn't gererate a core dump, i tried with different apps to force a dump and it worked but with nscd it just don't do it (or i'm doing it wrong). the output from debug level 10 is still the same. when i say apps doesn't work anymore it means they can't reach the host, i have to manually close the apps (firefox, kmail, etc (strangelly in the command line.. wget, ssh do work nicelly without nscd)), start nscd and start the apps. i've also noticed this happens more while using a junipper java vnp client that creates a tun device (tun0). but this happens more with the vpn (lets say twice an hour and without the vpn every two or three hours it happens). thanks a lot pedro
I'm also suffering from stability problems with nscd. The 30 systems run openSuSE 11.0 on 64 bit. I can't figure out any systematics yet, I start logging now.
*** This bug has been marked as a duplicate of bug 387202 ***