Hi,

This is partly a duplicate (sort of) of BZ 873382. It is sort of a duplicate because that bug is to get uwsgi to get rid of that hacky call in the systemd file. I pushed a build to stable just today, should arrive in -updates shortly, but if you want to try it before that happens here's the build:

http://koji.fedoraproject.org/koji/buildinfo?buildID=375043

Please let me know if this fixes the issue with uwsgi respawning.


Cheers

Alright, updated, currently not starting, however, also not running a default config (but a fairly standard config).

First error was:

Jan 25 09:37:47 TRS80 systemd[1]: Starting uWSGI Emperor Service...
Jan 25 09:37:47 TRS80 uwsgi[2421]: [uWSGI] getting INI configuration from /etc/uwsgi.ini
Jan 25 09:37:47 TRS80 uwsgi[2421]: *** Starting uWSGI 1.2.6 (64bit) on [Fri Jan 25 09:37:47 2013] ***
Jan 25 09:37:47 TRS80 uwsgi[2421]: compiled with version: 4.7.2 20121109 (Red Hat 4.7.2-8) on 27 December 2012 02:3
5:22
Jan 25 09:37:47 TRS80 uwsgi[2421]: detected number of CPU cores: 2
Jan 25 09:37:47 TRS80 uwsgi[2421]: current working directory: /
Jan 25 09:37:47 TRS80 uwsgi[2421]: writing pidfile to /run/uwsgi/uwsgi.pid
Jan 25 09:37:47 TRS80 uwsgi[2421]: detected binary path: /usr/sbin/uwsgi
Jan 25 09:37:47 TRS80 uwsgi[2421]: setgid() to 988
Jan 25 09:37:47 TRS80 uwsgi[2421]: setuid() to 991
Jan 25 09:37:47 TRS80 uwsgi[2421]: *** WARNING: you are running uWSGI without its master process manager ***
Jan 25 09:37:47 TRS80 uwsgi[2421]: your memory page size is 4096 bytes
Jan 25 09:37:47 TRS80 uwsgi[2421]: detected max file descriptor number: 1024
Jan 25 09:37:47 TRS80 uwsgi[2421]: lock engine: pthread robust mutexes
Jan 25 09:37:47 TRS80 uwsgi[2421]: *** starting uWSGI Emperor ***
Jan 25 09:37:47 TRS80 uwsgi[2421]: uwsgi socket 0 bound to UNIX address /run/uwsgi/uwsgi.socket fd 4
Jan 25 09:37:47 TRS80 uwsgi[2421]: your server socket listen backlog is limited to 100 connections
Jan 25 09:37:47 TRS80 uwsgi[2421]: *** Operational MODE: single process ***
Jan 25 09:37:47 TRS80 uwsgi[2421]: *** no app loaded. going in full dynamic mode ***
Jan 25 09:37:47 TRS80 uwsgi[2421]: *** uWSGI is running in multiple interpreter mode ***
Jan 25 09:37:47 TRS80 uwsgi[2421]: spawned uWSGI worker 1 (and the only) (pid: 2421, cores: 1)
Jan 25 09:37:48 TRS80 uwsgi[2421]: [emperor-tyrant] invalid permissions for file test.ini
Jan 25 09:37:52 TRS80 uwsgi[2421]: [emperor-tyrant] invalid permissions for file test.ini


Easy enough, chown uwsgi:uwsgi /etc/uwsgi.d/test.ini

Now, I receive:

Jan 25 10:59:56 TRS80 uwsgi[3900]: [emperor-tyrant] dropping privileges to 991 988 for instance test.ini
Jan 25 10:59:56 TRS80 uwsgi[3900]: setgroups(): Operation not permitted [emperor.c line 277]
Jan 25 10:59:56 TRS80 uwsgi[3900]: removed uwsgi instance test.ini

And a restart every second.

So somewhere along the line, uwsgi lost the ability to do setgroups().

This happens when using both systemctl or starting from the command line.

So I tried blowing away the current configs, and reinstalling (disabling and re-enabling the service).  Put back my configs, and still the same issue.

Current configs:

uwsgi.ini:

[root@TRS80 uwsgi.d]# cat /etc/uwsgi.ini 
[uwsgi]
uid = uwsgi
gid = uwsgi
pidfile = /run/uwsgi/uwsgi.pid
emperor = /etc/uwsgi.d
emperor-tyrant = true
master = true

[root@TRS80 uwsgi.d]# cat /etc/uwsgi.d/test.ini 
[uwsgi]
plugin = python
pythonpath = /var/www/
mount = /=hello:application
socket = 0.0.0.0:3031
logto = /var/log/uwsgi/uwsgi.log

If I set emperor-tyrant to false, I can start fine from the command line, starting from systemctl starts the service, but I still receive the timeout 90 seconds later and it restarts uwsgi over.  Command line shows:

[root@TRS80 uwsgi.d]# systemctl start uwsgi
Job for uwsgi.service failed. See 'systemctl status uwsgi.service' and 'journalctl -xn' for details.

[root@TRS80 uwsgi.d]# systemctl status uwsgi.service 
uwsgi.service - uWSGI Emperor Service
	  Loaded: loaded (/usr/lib/systemd/system/uwsgi.service; enabled)
	  Active: activating (start) since Fri 2013-01-25 21:16:31 CST; 31s ago
	 Process: 13105 ExecStartPre=/bin/chown uwsgi:uwsgi /run/uwsgi (code=exited, status=0/SUCCESS)
	 Process: 13102 ExecStartPre=/bin/mkdir -p /run/uwsgi (code=exited, status=0/SUCCESS)
	Main PID: 13108 (uwsgi)
	  CGroup: name=systemd:/system/uwsgi.service
		  ├─13108 /usr/sbin/uwsgi --ini /etc/uwsgi.ini
		  ├─13109 /usr/sbin/uwsgi --ini /etc/uwsgi.ini
		  ├─13110 /usr/sbin/uwsgi --ini test.ini
		  └─13111 /usr/sbin/uwsgi --ini test.ini

Jan 25 21:16:31 TRS80 uwsgi[13108]: setgid() to 988
Jan 25 21:16:31 TRS80 uwsgi[13108]: setuid() to 991
Jan 25 21:16:31 TRS80 uwsgi[13108]: your memory page size is 4096 bytes
Jan 25 21:16:31 TRS80 uwsgi[13108]: detected max file descriptor number: 1024
Jan 25 21:16:31 TRS80 uwsgi[13108]: lock engine: pthread robust mutexes
Jan 25 21:16:31 TRS80 uwsgi[13108]: *** Operational MODE: no-workers ***
Jan 25 21:16:31 TRS80 uwsgi[13108]: spawned uWSGI master process (pid: 13108)
Jan 25 21:16:31 TRS80 uwsgi[13108]: *** starting uWSGI Emperor ***
Jan 25 21:16:32 TRS80 uwsgi[13108]: *** has_emperor mode detected (fd: 7) ***
Jan 25 21:16:32 TRS80 uwsgi[13108]: [uWSGI] getting INI configuration from test.ini


Let me know if you need more info.

Thanks, this is useful... I don't think this is exactly related but a bug nonetheless. Let's close this one and track the setgroups() stuff in a different one, I'll poke the uwsgi mailing list for details.

All right so there's probably no need for a new bug, I asked around, you can see the response here:

http://lists.unbit.it/pipermail/uwsgi/2013-February/005370.html

tl;dr I shouldn't be putting "emperor-tyrant" in the emperor.ini. I pushed https://admin.fedoraproject.org/updates/uwsgi-1.2.6-7.fc18 which only has that change... 

diff --git a/emperor.ini b/emperor.ini
index 562b57b..9dfecca 100644
--- a/emperor.ini
+++ b/emperor.ini
@@ -4,4 +4,3 @@ gid = uwsgi
 socket = /run/uwsgi/uwsgi.socket
 pidfile = /run/uwsgi/uwsgi.pid
 emperor = /etc/uwsgi.d
-emperor-tyrant = true

And try again.