|Summary:||pam_cgroup is slow when cgrules.conf and /etc/passwd are large|
|Product:||Red Hat Enterprise Linux 6||Reporter:||Andy Grimm <agrimm>|
|Component:||libcgroup||Assignee:||Peter Schiffer <pschiffe>|
|Status:||CLOSED ERRATA||QA Contact:||Red Hat Kernel QE team <kernel-qe>|
|Version:||6.4||CC:||jeder, jgoulding, jkurik, jsafrane, jwest, mfisher, mmahut, paran+rhbugzilla, twiest, varekova, wduffee|
|Fixed In Version:||Doc Type:||Bug Fix|
Previously, the pam_cgoup PAM module did not use cache. Consequently, when a system had several thousand users and the cgrules.conf file contained several thousand lines of configuration settings, the login time could take several seconds. This was because the libcgroup code read /etc/passwd once for every line in cgrules.conf, unless the CGFLAG_USECACHE flag was used. With this update, the libcgroup code no longer reads the /etc/passwd file once for every line in cgrules.conf, and the login time is no longer affected in the described scenario.
|:||973353 (view as bug list)||Environment:|
|Last Closed:||2013-11-21 22:34:04 UTC||Type:||Bug|
|oVirt Team:||---||RHEL 7.3 requirements from Atomic Host:|
|Bug Depends On:|
|Bug Blocks:||973353, 974471|
Description Andy Grimm 2013-06-10 18:34:45 UTC
Description of problem: When a system has thousands of users and cgrules.conf has thousands of lines, this can add several seconds to login times. Version-Release number of selected component (if applicable): libcgroup-0.37-7.1.el6_4.x86_64 How reproducible: Always Steps to Reproduce: 1. Create 5000 users on a system 2. Create one or more lines per user in /etc/cgrules.conf ; for example: 513ac93f500446a08a000091 cpu,cpuacct,memory,net_cls,freezer /openshift/513ac93f500446a08a000091 3. Add "session optional pam_cgroup.so" to /etc/pam.d/sshd Actual results: Logins will take several seconds. Expected results: Login should take less than one second. Additional info: The root cause of this seems to be that the libcgroup code reads /etc/passwd once for every line in cgrules.conf , unless CGFLAG_USECACHE is used.
Comment 1 Andy Grimm 2013-06-10 18:36:04 UTC
Created attachment 759312 [details] Patch to use CGFLAG_USECACHE flag
Comment 3 Thomas Wiest 2013-06-10 18:55:15 UTC
This is currently affecting OpenShift Online. It would be great if it could be fixed quickly. Thanks!
Comment 7 Chris Williams 2013-06-11 19:16:45 UTC
*** Bug 973353 has been marked as a duplicate of this bug. ***
Comment 14 Mike Gahagan 2013-09-19 18:35:50 UTC
confirmed this is fixed in libcgroup-pam-0.40.rc1-3, reproduced with libcgroup-pam-0.37-7.1.el6_4.
Comment 15 errata-xmlrpc 2013-11-21 22:34:04 UTC
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. http://rhn.redhat.com/errata/RHBA-2013-1685.html
Comment 16 Pär Lindfors 2014-02-24 21:19:45 UTC
FYI: The released errata fix for this bug completely disables all functionality of pam_cgroup. This makes the reported slow performance go away, since doing nothing is a very fast operation. I have reported this as bug 1060227. Personally I don't think libcgroup have been designed for the reported use case. If you have thousands of users you probably need to avoid having thousands of rules in cgrules.conf, for example by using templates in that config file.