| Summary: | Service does not start after NetworkManager-wait-online.service | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | [Fedora] Fedora | Reporter: | Marcus Moeller <marcus.moeller> | ||||||
| Component: | NetworkManager | Assignee: | Dan Williams <dcbw> | ||||||
| Status: | CLOSED CURRENTRELEASE | QA Contact: | Fedora Extras Quality Assurance <extras-qa> | ||||||
| Severity: | unspecified | Docs Contact: | |||||||
| Priority: | unspecified | ||||||||
| Version: | 15 | CC: | dcbw, harald, jklimes, johannbg, kay, lpoetter, metherid, mschmidt, notting, plautrba | ||||||
| Target Milestone: | --- | ||||||||
| Target Release: | --- | ||||||||
| Hardware: | Unspecified | ||||||||
| OS: | Unspecified | ||||||||
| Whiteboard: | |||||||||
| Fixed In Version: | Doc Type: | Bug Fix | |||||||
| Doc Text: | Story Points: | --- | |||||||
| Clone Of: | Environment: | ||||||||
| Last Closed: | 2012-06-13 14:05:53 UTC | Type: | --- | ||||||
| Regression: | --- | Mount Type: | --- | ||||||
| Documentation: | --- | CRM: | |||||||
| Verified Versions: | Category: | --- | |||||||
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |||||||
| Cloudforms Team: | --- | Target Upstream Version: | |||||||
| Attachments: |
|
||||||||
|
Description
Marcus Moeller
2011-09-20 11:29:47 UTC
Created attachment 524043 [details]
dmesg.out
dmesg output with log_buf_len=1M systemd.log_target=kmsg systemd.log_level=debug set
Created attachment 524045 [details]
/var/log/messages during service startup
It seems to be related to nscd.service Stopping LSB: Starts the Name Switch Cache Daemon... has been logged during system startup. Disabling nscd (which is not a good idea in general) let's the other services start correctly. The wait-online service appears to have timed out. It has a default timeout of 30 seconds. I have already tried to set it to 60sec, but it only takes longer for the service to fail. Besides that, everything is fine, if I disable nscd, so I guess it has to be related to that one. Ok, it seems to work fine again, using these systemd service definitions:
# cat nscd.service
[Unit]
Description=Name Switch Cache Daemon
After=syslog.target network.target
[Service]
Type=forking
PIDFile=/run/nscd/nscd.pid
EnvironmentFile=-/etc/sysconfig/nscd
ExecStart=/usr/sbin/nscd $NSCD_OPTIONS
ExecReload=/usr/sbin/nscd -i passwd
ExecReload=/usr/sbin/nscd -i group
ExecReload=/usr/sbin/nscd -i hosts
ExecReload=/usr/sbin/nscd -i services
ExecStop=/usr/sbin/nscd -K
[Install]
WantedBy=multi-user.target
# cat /etc/tmpfiles.d/nscd.conf
d /run/nscd 0755 root root -
(In reply to comment #7) > Ok, it seems to work fine again, using these systemd service definitions: Nice to hear that. However, that's not the default service file, right? nscd is still using an initscript /etc/init.d/nscd. That should be fixed, I guess. (In reply to comment #6) > Besides that, everything is fine, if I disable nscd, so I guess it has to be > related to that one. There's a bug 700507 (RHEL 6) on nscd due to these lines: Sep 20 15:35:02 mymachine nscd: Can't send to audit system: USER_AVC avc: netlink poll: error 4#012: exe="?" sauid=28 hostname=? addr=? terminal=? Sep 20 15:35:02 mymachine nscd: Can't send to audit system: USER_AVC avc: netlink recvfrom: error 1#012: exe="?" sauid=28 hostname=? addr=? terminal=? Sep 20 15:35:02 mymachine nscd: Can't send to audit system: USER_AVC avc: netlink thread: errors encountered, terminating#012: exe="?" sauid=28 hostname=? addr=? terminal=? It's fixed in RHEL 6, however the fix is not delivered to Fedora 15 yet. Looking into bodhi, the fix is included in glibc-2.14.90-9 (F16) that is in 'testing' state, no valid update for F15. I'm not sure if the bug is connected with the issue here. From NM point of view, there is a big delay (29sec!) between these two lines: Sep 20 15:35:06 mymachine NetworkManager[889]: <info> Activation (eth0) Stage 5 of 5 (IP Configure Commit) started... Sep 20 15:36:35 mymachine NetworkManager[889]: <info> Activation (eth0) Stage 5 of 5 (IP Configure Commit) complete. even if they are logged in the same function: src/nm-device.c:nm_device_activate_stage5_ip_config_commit() nscd somehow badly interferes with NM, probably. It would be useful to find out where NM is stuck. E.g. with: sudo pstack `pidof NetworkManager` Also appending '--log-level=DEBUG' to ExecStart in /lib/systemd/system/NetworkManager.service will help increasing logging level of NM. Works in more recent Fedora versions |