Broken systemd unit for gnocchi-api

[Daniele Venzano]

gnocchi-api is a developer wrapper arount a test server.
The Gnocchi documentation (http://gnocchi.xyz/operating.html) suggests using Apache WSGI to run it. The Ubuntu packages should contain the Apache configuration like is done for Keystone and Cinder.

Anyway the systemd unit is broken because gnocchi-api is called with the wrong options:

Jul 31 14:34:00 bfeb systemd[1]: Started OpenStack Gnocchi Api.
Jul 31 14:34:02 bfeb gnocchi-api[10203]: usage: gnocchi-api [-h] [--port PORT] -- [passed options]
Jul 31 14:34:02 bfeb gnocchi-api[10203]: gnocchi-api: error: unrecognized arguments: --config-file=/etc/gnocchi/gnocchi.conf --log-file=/var/log/gnocchi/gnocchi-api.log
Jul 31 14:34:02 bfeb systemd[1]: gnocchi-api.service: Main process exited, code=exited, status=2/INVALIDARGUMENT
Jul 31 14:34:02 bfeb systemd[1]: gnocchi-api.service: Unit entered failed state.
Jul 31 14:34:02 bfeb systemd[1]: gnocchi-api.service: Failed with result 'exit-code'.
Jul 31 14:34:02 bfeb systemd[1]: gnocchi-api.service: Service hold-off time over, scheduling restart.
Jul 31 14:34:02 bfeb systemd[1]: Stopped OpenStack Gnocchi Api.
Jul 31 14:34:02 bfeb systemd[1]: gnocchi-api.service: Start request repeated too quickly.
Jul 31 14:34:02 bfeb systemd[1]: Failed to start OpenStack Gnocchi Api.

This is Ubuntu Xenial with the Ocata cloud archive for openstack.

[Sumit KUmar (sumit.kumar)]

I am having the same problem:

* gnocchi-api.service - OpenStack Gnocchi Api
   Loaded: loaded (/lib/systemd/system/gnocchi-api.service; disabled; vendor preset: enabled)
   Active: inactive (dead)

Sep 02 04:53:57 neutron1 gnocchi-api[19591]: usage: gnocchi-api [-h] [--port PORT] -- [passed options]
Sep 02 04:53:57 neutron1 gnocchi-api[19591]: gnocchi-api: error: unrecognized arguments: --config-file=/etc/gnocchi/gnocchi.conf --log-file=/var/log/gnocchi/gnocchi-api.log
Sep 02 04:53:57 neutron1 systemd[1]: gnocchi-api.service: Main process exited, code=exited, status=2/INVALIDARGUMENT
Sep 02 04:53:57 neutron1 systemd[1]: gnocchi-api.service: Unit entered failed state.
Sep 02 04:53:57 neutron1 systemd[1]: gnocchi-api.service: Failed with result 'exit-code'.
Sep 02 04:53:57 neutron1 systemd[1]: gnocchi-api.service: Service hold-off time over, scheduling restart.
Sep 02 04:53:57 neutron1 systemd[1]: Stopped OpenStack Gnocchi Api.
Sep 02 04:53:57 neutron1 systemd[1]: gnocchi-api.service: Start request repeated too quickly.
Sep 02 04:53:57 neutron1 systemd[1]: Failed to start OpenStack Gnocchi Api.
Sep 04 16:24:03 neutron1 systemd[1]: Stopped OpenStack Gnocchi Api.

Is there any workaround for this?

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[KI4RBC (mikemcdonough)]

To add a little more of the back story:

/etc/os_release:

NAME="Ubuntu"
VERSION="16.04.4 LTS (Xenial Xerus)"
ID=ubuntu
ID_LIKE=debian
PRETTY_NAME="Ubuntu 16.04.4 LTS"
VERSION_ID="16.04"
HOME_URL="http://www.ubuntu.com/"
SUPPORT_URL="http://help.ubuntu.com/"
BUG_REPORT_URL="http://bugs.launchpad.net/ubuntu/"
VERSION_CODENAME=xenial
UBUNTU_CODENAME=xenial

Package:

gnocchi-api/xenial-updates,now 3.1.2-0ubuntu1~cloud0 all [installed]

Errors in syslog:
Mar 6 03:01:25 pod2controller-admin gnocchi-api[100950]: usage: gnocchi-api [-h] [--port PORT] -- [passed options]
Mar 6 03:01:25 pod2controller-admin gnocchi-api[100950]: gnocchi-api: error: unrecognized arguments: --config-file=/etc/gnocchi/gnocchi.conf --log-file=/var/log/gnocchi/gnocchi-api.log

The comment in `/etc/init.d/gnocchi-api` seems to suggest that this init script is auto-generated.

# The content after this line comes from openstack-pkg-tools
# and has been automatically added to a .init.in script, which
# contains only the descriptive part for the daemon. Everything
# else is standardized as a single unique script.

So this standardized script is setting $DAEMON_ARGS with the above arguments while `/usr/bin/gnocchi-api` is expecting only a port number to listen on.

I am able to execute `/etc/bin/gnocchi-api` and sucessfully test the api

[root@controller gnocchi(admin)]$ /usr/bin/gnocchi-api --port 8041
********************************************************************************
STARTING test server gnocchi.rest.app.build_wsgi_app
Available at http://10.3.35.40:8041/
DANGER! For testing only, do not use in production
********************************************************************************
192.168.100.40 - - [06/Mar/2018 22:26:53] "GET / HTTP/1.1" 200 150

[root@controller gnocchi(admin)]$ api-test.sh -a gnocchi -c controller-int -d
...
Mar 06 22:26:53 pod2controller /usr/local/bin/api-test.sh: INFO: gnocchi test passed on http://pod2controller-int:8041/
{
    "versions": [
        {
            "id": "v1.0",
            "links": [
                {
                    "href": "http://controller-int:8041/v1/",
                    "rel": "self"
                }
            ],
            "status": "CURRENT",
            "updated": "2015-03-19"
        }
    ]
}
[root@controller gnocchi(admin)]$

I also am able to use uwsgi with this ini file
[uwsgi]
http = 0.0.0.0:8041
wsgi-file = /usr/bin/gnocchi-api
master = true
die-on-term = true
threads = 32
processes = 32
enabled-threads = true
thunder-lock = true
plugins = python
buffer-size = 65535
lazy-apps = true
add-header = Connection: close

Perhaps @zigo can chime in here?

[KI4RBC (mikemcdonough)]

as a work-around I did the following:

mv /usr/bin/gnocchi-api /usr/bin/gnocchi-wsgi-api
cat << 'EOF' >> /usr/bin/gnocchi-api
#!/bin/sh
GID=$(/usr/bin/id -g gnocchi)
/usr/local/bin/uwsgi /etc/gnocchi/uwsgi.ini --gid $GID
EOF
cat << 'EOF' >> /etc/gnocchi/uwsgi.ini
[uwsgi]
http = 0.0.0.0:8041
wsgi-file = /usr/bin/gnocchi-wsgi-api
master = true
die-on-term = true
threads = 32
processes = 32
enabled-threads = true
thunder-lock = true
plugins = python
buffer-size = 65535
lazy-apps = true
add-header = Connection: close
EOF
service gnocchi-api restart

[KI4RBC (mikemcdonough)]

My team has a suspicion adding uwsgi package is causing a problem with openstack-dashboard. I am looking to patch the init script that calls the original gnocchi-api to call the correct arguments this week and will post it here.

[KI4RBC (mikemcdonough)]

So scratch my earlier workaround or patching. After reading what /etc/init/gnocchi-api.conf is doing around lines 31 to 36 I found I can modify startup behavior by creating `/etc/defaults/gnocchi-api` and adding the following contents

```bash
DAEMON_ARGS="--port 8041 -- --log-file=/var/log/gnocchi/gnocchi-api.log"
CONFIG_FILE="/etc/gnocchi/gnocchi.conf"
USE_SYSLOG="no"
USE_LOGFILE="yes"
```

It seems the extra "--" is not being passed to the start-stop-daemon function

Now my only problem is that API requests are still showing up in `/var/log/syslog` even though this is not what I see in the arguments reported by systemctl status

```bash
   CGroup: /system.slice/gnocchi-api.service
           └─110578 /usr/bin/python2.7 /usr/bin/gnocchi-api --port 8041 -- --log-file=/var/log/gnocchi/gnocchi-api.log
```

[Andreas (fattony666)]

It seems that the latest version (4.2.0) does not have the gnocchi-api service unit installed.

Failed to restart gnocchi-api.service: Unit gnocchi-api.service not found.