Jakub Thanks for those repos. That was MUCH easier to get installed.
Things seem much faster now as well, everything seems to be running pretty well. if I turn the ignore users off login times are too long and I get disconnected, but logging in again gets me in and groups have members, how long does group member info stay in the cache? Haven't decided yet if I want to leave that on or off yet. Probably off if the cache expires and it disconnects everybody every login the first login. That would get tiresome. I am able now to get a consistent list of groups for the users I tried. There were some differences between what 'id' returned and what was in AD Users & Computers, but I tracked those down to either being groups with Category 'Distribution List' and not 'Security', or groups in built-in containers. I went looking for one of those specifically in the logs and SSSD does in fact find it, but it explains what's up with it and why it's not included in 'id': (Thu Feb 4 10:55:26 2016) [sssd[be[austin.utexas.edu]]] [sdap_get_primary_name] (0x0400): Processing object Pre-Windows 2000 Compatible Access (Thu Feb 4 10:55:26 2016) [sssd[be[austin.utexas.edu]]] [sdap_add_incomplete_groups] (0x1000): Mapping group [Pre-Windows 2000 Compatible Access] objectSID to unix ID (Thu Feb 4 10:55:26 2016) [sssd[be[austin.utexas.edu]]] [sdap_add_incomplete_groups] (0x2000): Group [Pre-Windows 2000 Compatible Access] has objectSID [S-1-5-32-554] (Thu Feb 4 10:55:26 2016) [sssd[be[austin.utexas.edu]]] [sdap_idmap_sid_to_unix] (0x0400): Object SID [S-1-5-32-554] is a built-in one. (Thu Feb 4 10:55:26 2016) [sssd[be[austin.utexas.edu]]] [sdap_add_incomplete_groups] (0x2000): Group [Pre-Windows 2000 Compatible Access] cannot be mapped. Treating as a non-POSIX group Thanks for descriptive logs, they seem to be a luxury. On my RHEL 6 vm with the preview installed, I've commented out my cron job to renew the machine ticket. Is anything else required to have that happen or does > krb5_renewable_lifetime = 7d > krb5_renew_interval = 6h cover machine tickets the same as user tickets? As it turns out I just happened to be tailing the log file and saw the following. Does adcli need to be a particular version to have the update option? When does the update try to happen, or am I just lucky I caught it? I see it's once a day, does it try after service startup then schedules for each day after that? (Fri Feb 5 14:08:54 2016) [sssd[be[austin.utexas.edu]]] [be_ptask_execute] (0x0400): Task [AD machine account password renewal]: executing task, timeout 60 seconds (Fri Feb 5 14:08:54 2016) [sssd[be[austin.utexas.edu]]] [child_handler_setup] (0x2000): Setting up signal handler up for pid [51099] (Fri Feb 5 14:08:54 2016) [sssd[be[austin.utexas.edu]]] [child_handler_setup] (0x2000): Signal handler set up for pid [51099] (Fri Feb 5 14:08:54 2016) [sssd[be[austin.utexas.edu]]] [child_sig_handler] (0x1000): Waiting for child [51099]. (Fri Feb 5 14:08:54 2016) [sssd[be[austin.utexas.edu]]] [child_sig_handler] (0x0020): child [51099] failed with status [2]. (Fri Feb 5 14:08:54 2016) [sssd[be[austin.utexas.edu]]] [read_pipe_handler] (0x0400): EOF received, client finished (Fri Feb 5 14:08:54 2016) [sssd[be[austin.utexas.edu]]] [ad_machine_account_password_renewal_done] (0x1000): --- adcli output start--- adcli: 'update' is not a valid adcli command. See 'adcli --help' ---adcli output end--- (Fri Feb 5 14:08:54 2016) [sssd[be[austin.utexas.edu]]] [be_ptask_done] (0x0400): Task [AD machine account password renewal]: finished successfully (Fri Feb 5 14:08:54 2016) [sssd[be[austin.utexas.edu]]] [be_ptask_schedule] (0x0400): Task [AD machine account password renewal]: scheduling task 86400 seconds from last execution time [1454789334] Just realized this is password, not ticket. Nevermind. Still nice to see though. I tried searching the log for 'kinit' but only found, I think, the kinit's for the users logging in, but I'm not sure which context it's initing. Looks to be the computer, I never see any mention of any users kiniting however. We'll see if my user tickets get renewed later tonight. The troubleshooting page was helpful, I'm not sure how I missed that one but it helps. Once again, thanks for the time. Todd -----Original Message----- From: Jakub Hrozek [mailto:[email protected]] Sent: Friday, February 5, 2016 2:56 AM To: [email protected] Subject: [SSSD-users] Re: I'm new to everything! On Thu, Feb 04, 2016 at 09:38:00PM +0000, Mote, Todd wrote: > Hello all > > I'm a Windows Domain Admin where I work and am working on using SSSD to get > some identity management consistency to our Linux RHEL 6 and 7 fleet, a long > overdue state. > > I've gotten pretty far, we're using the build that's been released from Red > Hat, 1.12.4-47 on RHEL 6 and 1.13.0-40 on RHEL 7. I've joined them both with > adcli since realmd isn't available on RHEL 6. I'm trying to keep things > consistent between OS releases. I can log in, process group policy, even ssh > using Kerberos (which I found something weird concerning character case in > the ticket cache with, but I think that's more Kerberos and adcli and ssh, > than sssd). It's been a fun adventure for this Windows guy. I've learned a > lot about Linux through this process. Hopefully, this email isn't so long > that it wears anybody out. > > I have come across some things I wanted to get some advice about. Our > AD is VERY large. On the order of 7 million user accounts at this > point. I've had to overcome some permission issues in AD, using a > machine keytab, SASL and GSSAPI for lookups meant Domain Computers had > to have rights to read all of the necessary attributes on users. We > have some FERPA and HIPPA issues to deal with, so a general Domain > Computers - Read permission won't work, and it seems that > Authenticated Users isn't processed quite right for Computer objects. > However, I got that worked out but turning up the logging to 9 and > seeing what attributes you are looking for from users. I applied > Domain Computers - Read to just those attributes and it was enough. > But am now having some problems getting the right groups from users > after they log in. After lots of trial and error, I arrived at the > following sssd.conf which works pretty well. we have a single forest, > single domain AD, and a security office that cringes at any number > anywhere that has 9 digits in it. (in case you were wondering about > the idmap range) > > [sssd] > config_file_version = 2 > debug_level = 9 > domains = austin.utexas.edu > services = nss, pam, pac > > [nss] > > [pam] > > [pac] > > [domain/austin.utexas.edu] > debug_level = 9 > id_provider = ad > access_provider = ad > ad_domain = austin.utexas.edu > ad_server = dc01.austin.utexas.edu > auth_provider = ad > cache_credentials = true > ldap_schema = ad > ldap_idmap_range_min = 1000000000 > ldap_idmap_range_size = 20000000 > ldap_idmap_default_domain = austin.utexas.edu > ldap_idmap_default_domain_sid = <ourdomainSID> > override_homedir = /home/AUSTIN/%u > default_shell = /bin/bash > krb5_use_enterprise_principal = true > krb5_renewable_lifetime = 7d > krb5_renew_interval = 6h > krb5_realm = AUSTIN.UTEXAS.EDU > # ad_gpo_access_control = permissive > ad_gpo_access_control = enforcing > ad_gpo_cache_timeout = 5 > dyndns_update = true > dyndns_update_ptr = false > dyndns_refresh_interval = 86400 > dyndns_ttl = 3600 > ignore_group_members = true > ldap_use_tokengroups = false > ldap_group_nesting_level = 0 > > I ended up at ignore_group_members=true because Domain Users has LOTS > of users in it, as do other programmatically populated groups, not > using token groups and setting nesting level to 0 and am very close to > replicating what I see on the memberof tab in ADU&C for the user object. > I was having LDAP search time outs due to size and group enumeration. > But the groups returned by 'id' are not quite complete and seems to > change between cache clears and service restarts. I was reading about > 1.13.3 and the closed ticket about flaky group memberships and > ignore_group_members and thought I might give it a try. I think this must be https://bugzilla.redhat.com/show_bug.cgi?id=1212610 -- which was backported to RHEL-6.7's 1.12 version. Nonetheless, it would be great to get some test results with packages from the repos below. > Though I'm finding that to be a > lot harder than I thought it would be. I downloaded the source from > https://fedorahosted.org/released/sssd/ and unpacked it, but I'm not > sure of where to go from here, so I looked for an rpm, because I do at > least know how to yum install, but ran into a tangly mess of dependencies. > > I guess my questions are thus: > Are there any instructions for the weak linux skilled windows admin to get > 1.13.3 installed without a lot of trouble? I looked in the BUILD.txt in the > package and it lists https://fedorahosted.org/sssd/wiki/DevelTutorials as a > place for instructions, but the link doesn't go anywhere. The readme had > this mailing list. So here I am. I have a test repo with the packages we're planning for 6.8: https://copr.fedorainfracloud.org/coprs/jhrozek/SSSD-6.8-preview/ and Lukas maintains a repo which tracks the upstream releases which also includes EPEL-7 builds: https://copr.fedorainfracloud.org/coprs/g/sssd/sssd-1-13/ If these don't help, it would be nice to describe the missing group memberships and look into the logs for details. Hopefully the troubleshooting page would help: https://fedorahosted.org/sssd/wiki/Troubleshooting > > Second are there any general best practices with sssd and AD anywhere? > I've blindly just come across stuff, like krb5_renew_interval for user > ticket renewal, our machines were falling off the domain after 7 days, > so we also now have a cron job that runs every so often to keep the > computer's ticket refreshed. (I see that is a RFE though!) Yes, this one should be fixed in the 6.8 preview repo at least. Because the ticket was only fixed after 1.13.3 was released, this feature is not in the upstream repo yet, we only cherry-picked the patches to RHEL. > > I could go on, but this is long enough, hopefully no one will throw > tomatoes. :) Thanks in advance for any time you spend on this. SSSD > will solve so many issues for us if we can get it working reliably here. > I even have a colleague working on a puppet module for joining > machines to AD at build time! Great, this would be nice to see! _______________________________________________ sssd-users mailing list [email protected] https://lists.fedorahosted.org/admin/lists/[email protected] _______________________________________________ sssd-users mailing list [email protected] https://lists.fedorahosted.org/admin/lists/[email protected]
