** Description changed: [Impact] sssd can switch to an offline mode of operation when it cannot reach the authentication or id backend. It uses several methods to assess the situation, and one of them is monitoring the /etc/resolv.conf file for changes. In ubuntu that file is a symlink to /run/systemd/resolve/stub- resolv.conf, but the target doesn't exist at all times during boot. It's expected that symlink to be broken for a while during boot. Turns out that the monitoring that sssd was doing on /etc/resolv.conf didn't take into consideration that what could change was the *target* of the symlink. it completely ignored that fact, and didn't notice when the resolv.conf contents actually changed in this scenario, which resulted in sssd staying in the offline mode when it shouldn't. There are two fixes being pulled in for this SRU: a) fix the monitoring of the target of the /etc/resolv.conf symlink b) change the fallback polling code to keep trying, instead of giving up right away [Test Case] It's recommended to test this in a lxd container, or a vm. Preparation steps. When prompted for an openldap/slapd password, chose any password you want. It won't be needed again: $ sudo apt update && sudo apt dist-upgrade $ sudo apt install sssd-ldap sssd-tools sssd-dbus slapd dnsmasq Become root: $ sudo su - Detect your ip: # export interface=$(ip route | grep default | sed -r 's,^default via .* dev ([a-z0-9]+) .*,\1,') # export ip=$(ip addr show dev $interface | grep "inet [0-9]" | awk '{print $2}' | cut -d / -f 1) Confirm the $ip variable is correct for your case: # echo $ip Create /etc/dnsmasq.d/sssd-test.conf using your real ip: # cat > /etc/dnsmasq.d/sssd-test.conf <<EOF host-record=ldap01.example.com,$ip listen-address=$ip interface=$interface except-interface=lo - bind-interface + bind-interfaces EOF restart dnsmasq # systemctl restart dnsmasq a) inotify test Create /etc/sssd/sssd.conf: # cat > /etc/sssd/sssd.conf <<EOF [sssd] config_file_version = 2 services = nss, pam, ifp domains = LDAP debug_level = 6 [domain/LDAP] id_provider = ldap ldap_uri = ldap://ldap01.example.com cache_credentials = True ldap_search_base = dc=example,dc=com EOF # chmod 0600 /etc/sssd/sssd.conf # rm /etc/resolv.conf # ln -s /etc/resolv.conf.target /etc/resolv.conf create good resolv.conf: # echo "nameserver $ip" > /etc/resolv.conf.good Confirm /etc/resolv.conf is a broken symlink: # ll /etc/resolv.conf* lrwxrwxrwx 1 root root 23 May 13 20:48 /etc/resolv.conf -> /etc/resolv.conf.target -rw-r--r-- 1 root root 24 May 13 20:48 /etc/resolv.conf.good Open another terminal/screen and tail the sssd logs with a grep: # tail -F /var/log/sssd/sssd.log | grep -i resolv Start sssd # systemctl restart sssd The tail output from that other terminal should say sssd is monitoring /etc/resolv.conf (that's the broken symlink): (Mon May 18 17:32:34 2020) [sssd] [_snotify_create] (0x0400): Added a watch for /etc/resolv.conf with inotify flags 0x8D88 internal flags 0x1 using function resolv_conf_inotify_cb after delay 1.0 Repeat this sssctl call until it shows the offline mode persistently: # sssctl domain-status LDAP Online status: Offline Active servers: LDAP: not connected Discovered LDAP servers: - ldap01.example.com "Unbreak" the symlink: # cp /etc/resolv.conf.good /etc/resolv.conf.target The log should now say: (Mon May 18 17:33:30 2020) [sssd] [process_dir_event] (0x0400): Not interested in resolv.conf.target This shows that sssd didn't pick up that resolv.conf changed, via the target of the symlink. Run sssctl again, and the online status will remain offline. With the fixed packages, the sssd startup log will say: (Mon May 18 17:17:06 2020) [sssd] [_snotify_create] (0x0400): Added a watch for /etc/resolv.conf.target with inotify flags 0x8D88 internal flags 0x1 using function resolv_conf_inotify_cb after delay 1.0 Showing that it's monitoring the symlink target. And after fixing the broken symlink, it will say: (Mon May 18 17:18:06 2020) [sssd] [process_dir_event] (0x0400): received notification for watched file [resolv.conf.target] under /etc Run sssctl again, it should almost immediately switch to online: # sssctl domain-status LDAP Online status: Online Active servers: LDAP: ldap01.example.com Discovered LDAP servers: - ldap01.example.com b) polling test Repeat the previous test, but with "try_inotify = false" in sssd.conf, like this: # cat > /etc/sssd/sssd.conf <<EOF [sssd] config_file_version = 2 services = nss, pam, ifp domains = LDAP debug_level = 6 try_inotify = false [domain/LDAP] id_provider = ldap ldap_uri = ldap://ldap01.example.com cache_credentials = True ldap_search_base = dc=example,dc=com EOF Then follow the other steps of test case (a). Upon startup, the sssd log will show: (Mon May 18 17:57:04 2020) [sssd] [monitor_config_file_fallback] (0x0080): file [/etc/resolv.conf] is missing. Will not update online status based on watching the file Note how it says it will not watch it for updates. With the fixed package, the log will show this instead: (Mon May 18 18:02:56 2020) [sssd] [monitor_config_file_fallback] (0x0020): file [/etc/resolv.conf] is missing. Will try again later. Notice how it says it will try again, and multiple times if you keep watching it (once every 10s). "Unbreak" the symlink: # cp /etc/resolv.conf.good /etc/resolv.conf.target And the log output will eventually show this: (Mon May 18 18:04:06 2020) [sssd] [monitor_config_file_fallback] (0x0020): file [/etc/resolv.conf] is missing. Will try again later. (Mon May 18 18:04:17 2020) [sssd] [signal_res_init] (0x0040): Reloading Resolv.conf. And sssctl will show the now online status: # sssctl domain-status LDAP Online status: Online Active servers: LDAP: ldap01.example.com Discovered LDAP servers: - ldap01.example.com [Regression Potential] Breaking sssd can mean preventing logins of network remote users on a system, but should still allow local users (and root) to login. Another possible source of regressions, but not related to the code changes in this update, if there are pre-existing incorrect changes to sssd.conf, they might prevent the service from restarting when this update is applied. [Other Info] Not at this time. [Original Description] SSSD 1.15.3-2ubuntu1 on 17.10/artful (previous versions on artful were also affected) is offline on boot and seems to stay offline forever (I waited over 20 minutes). sssd_nss.log: (Fri Oct 13 09:49:50 2017) [sssd[nss]] [sss_dp_get_reply] (0x0010): The Data Provider returned an error [org.freedesktop.sssd.Error.DataProvider.Offline] (Fri Oct 13 09:49:51 2017) [sssd[nss]] [sss_dp_get_reply] (0x0010): The Data Provider returned an error [org.freedesktop.sssd.Error.DataProvider.Offline] (Fri Oct 13 09:49:51 2017) [sssd[nss]] [sss_dp_get_reply] (0x0010): The Data Provider returned an error [org.freedesktop.sssd.Error.DataProvider.Offline] (Fri Oct 13 09:49:51 2017) [sssd[nss]] [sss_dp_get_reply] (0x0010): The Data Provider returned an error [org.freedesktop.sssd.Error.DataProvider.Offline] ... SSSD immediately returns to normal operation after restarting it or after sending SIGUSR2. A workaround for the problem is creating the file /etc/systemd/system/sssd.service.d/override.conf with contents [Unit] Requires=network-online.target After=network-online.target
-- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1723350 Title: sssd offline on boot, stays offline forever To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/sssd/+bug/1723350/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs