Bug 40371

Summary:	server_password_change stuck at restarting nscd
Product:	UCS	Reporter:	Michael Grandjean <grandjean>
Component:	General	Assignee:	Lukas Oyen <oyen>
Status:	CLOSED FIXED	QA Contact:	Arvid Requate <requate>
Severity:	normal
Priority:	P5	CC:	gohmann, requate
Version:	UCS 4.0	Flags:	oyen: Patch_Available+
Target Milestone:	UCS 4.2-2-errata
Hardware:	Other
OS:	Linux
What kind of report is it?:	---	What type of bug is this?:	---
Who will be affected by this bug?:	---	How will those affected feel about the bug?:	---
User Pain:		Enterprise Customer affected?:	Yes
School Customer affected?:		ISV affected?:
Waiting Support:		Flags outvoted (downgraded) after PO Review:
Ticket number:		Bug group (optional):
Max CVSS v3 score:
Attachments:	Move cache invalidation after daemon restart

Description Michael Grandjean

2016-01-05 22:10:39 CET

root@gydwag-opsi:~# ucr search --brief ^version
version/erratalevel: 342
version/patchlevel: 2
version/releasename: Walle
version/version: 4.0

server/role: memberserver

Installed 3rd party applications: opsi4ucs

> root@gydwag-opsi:~# ps auxf | grep '[n]scd'
> root      9707  0.0  0.0   4188   468 ?        S     2015   0:00                  \_ /bin/sh /usr/lib/univention-server/server_password_change.d/univention-nscd postchange
> root      9708  0.0  0.0   4188  1308 ?        S     2015   0:00                      \_ /bin/sh /etc/init.d/nscd restart
> root      9711  0.0  0.0  10588   768 ?        S     2015   0:00                          \_ /usr/sbin/nscd --invalidate passwd
> root      9895 27.8  4.4 540396 179432 ?       Rsl   2015 15211:18 /usr/sbin/nscd


Last lines of server_password_change.log on that day:

> [...]
> run-parts: executing /usr/lib/univention-server/server_password_change.d/univention-libnss-ldap postchange
> W: failed to convert the username messagebus to the uid
> File: /etc/libnss-ldap.conf
> run-parts: executing /usr/lib/univention-server/server_password_change.d/univention-nscd postchange
> Restarting Name Service Cache Daemon: nscd

Next day:
> Starting server password change (Tue Dec 22 01:02:04 CET 2015)
> No server password change scheduled for today, terminating without a change

Comment 1 Michael Grandjean

2016-01-05 22:13:26 CET

Workaround: Kill the oldest running nscd process (PID 9895 in the example above). "univention-nscd postchange" will then finish successfully.

Comment 2 Arvid Requate

2016-01-06 21:24:15 CET

IMHO we definitely should make /etc/init.d/nscd restart more robust against this.

It does  nscd --invalidate $table   and then stop_nscd. It must not hang in the first steps of this, ever. We should assure that it gets restarted reliably.

Comment 3 Michael Grandjean

2016-02-04 20:51:32 CET

2016012621000729 - same customer, other server

Comment 4 Lukas Oyen

2016-12-13 16:32:45 CET

Created attachment 8304 [details]
Move cache invalidation after daemon restart

Comment 5 Florian Best

2017-06-28 14:53:01 CEST

There is a Customer ID set so I set the flag "Enterprise Customer affected".

Comment 6 Lukas Oyen

2017-08-01 17:20:02 CEST

Committed in r81656 - r81657 (advisory r81658).

Comment 7 Arvid Requate

2017-08-28 19:52:07 CEST

Ok, works.

Comment 8 Arvid Requate

2017-09-13 16:35:04 CEST

<http://errata.software-univention.de/ucs/4.2/161.html>