Frontend stopped connecting to recorder. about frontend

[plague-doctor]

Oh, it is just easier for me to test different things.

ENTRYPOINT starts the container with parameters that cannot be overridden with CLI parameters.
CMD allows to do so.

In other words, when you specify:

ENTRYPOINT=/usr/bin/my-script.sh

then you cannot start your container with different parameter/script for example:

docker run -it my-docker-image /usr/bin/different-script.sh

The above example will start your container with /usr/bin/my-script.sh regardless what you specify as a starting parameter. In fact /usr/bin/different-script.sh will be interpreted as a parameter for your entrypoint.

When you use

CMD=["/usr/bin/my-script.sh"]

instead, you can override the starting point, so the above command will start your container with /usr/bin/different-script.sh.

When you specify both ENTRYPOINT and CMD, then whatever you put into the CMD becomes a parameter for your entrypoint. For example:

ENTRYPOINT=/usr/bin/my-script.sh
CMD=["parameter1", "parameter2"]

will be interpreted as:

/usr/bin/my-script.sh parameter1 parameter2

and you can override parameters with:

docker run -it my-docker-image parameter3

which will be corresponding to:

/usr/bin/my-script.sh parameter3

I hope this makes sense.

In my Dockerfile it can easily be moved to ENTRYPOINT. No problems with that ;-)

[jpmens]

Thank you.

Closing this now, as it's been answered in owntracks/recorder#388

[linusg]

Curious - which recorder version did you previously run? I just updated my recorder here to latest and it still seems to work fine, but I'm using a slightly different (also traefik) setup.

[plague-doctor]

I've got :latest as well, however when it started to fail, I have done some investigation ;-)
According to https://hub.docker.com/r/owntracks/recorder/tags it was 0.8.6-12. It has been updated 6 days ago to version 0.8.8...

[jpmens]

I don't think we've changed anything which would cause this, @plague-doctor. What do you get when you invoke

curl http://a.b.c.d:8083/api/0/version

against the Recorder running in your container?

[plague-doctor]

Ok, it looks like Recorder is binding to 127.0.0.1 instead to 0.0.0.0:

# netstat -ntlp
tcp        0      0 127.0.0.1:8083          0.0.0.0:*               LISTEN      13/ot-recorder

so I can nicely get response curl http://127.0.0.1:8083/api/0/version from within the container:

{"version":"0.8.8","git":"0.8.8-0-g0f86aa3af8"}

but not from other container (in my case from otr-frontend).
Recorder keeps ignoring env. variable OTR_HTTPHOST=0.0.0.0...

ot-recorder[9]: version 0.8.8 starting with STORAGEDIR=/store
ot-recorder[9]: connecting to MQTT ...
ot-recorder[9]: HTTP listener started on 127.0.0.1:8083

[jpmens]

There's a OTR_HTTPHOST in recorder.conf in the container. What is that set to?

[plague-doctor]

/config/recorder.conf is pretty much empty, except:

OTR_TOPICS="owntracks/#"

I can see it is set via env variables:

OTR_PASS=<edited>
OTR_HOST=<edited>
OTR_USER=<edited>
OTR_HTTPHOST=0.0.0.0
OTR_STORAGEDIR=/store
OTR_TOPIC=owntracks/#

but it is ignored...

I have noticed there is another version released (0.9.0) but the behavior is the same.

ot-recorder[10]: version 0.9.0 starting with STORAGEDIR=/store
ot-recorder[10]: connecting to MQTT 
ot-recorder[10]: HTTP listener started on 127.0.0.1:8083
ot-recorder[10]: Using storage at /store with precision 7
ot-recorder[10]: Subscribing to owntracks/# (qos=2)

[plague-doctor]

It is misbehaving due to wrong entrypoint in the Docker file. Currently there entrypoint is:

#!/bin/sh

if ! [ -f ${OTR_STORAGEDIR}/ghash/data.mdb ]; then
    ot-recorder --initialize
fi

ot-recorder ${OTR_TOPIC}

which gives me the issue. The problematic bit is:

ot-recorder ${OTR_TOPIC}

which should be rather:

ot-recorder --http-host ${OTR_HTTPHOST:-0.0.0.0} --http-port ${OTR_HTTPPORT:-8083} $OTR_OPTIONS ${OTR_TOPICS:-owntracks/#}

I have came up with my Docker image (plaguedr/owntracks-recorder) which works as expected. In case you are interested, here it is:

Dockerfile:

FROM owntracks/recorder
LABEL maintainer="Plague Doctor"
LABEL name="owntracks-recorder"
LABEL architecture="x86_64"
COPY start.sh /usr/bin/start.sh
RUN apk update && apk add mosquitto-clients && \
    chmod 775 /usr/bin/start.sh
CMD ["start.sh"]
ENTRYPOINT []

and start.sh:

#!/bin/sh

while true; do
    if mosquitto_sub -h "$OTR_HOST" -u "$OTR_USER" -P "$OTR_PASS" -p 1883 -t '$SYS/broker/uptime' -C 1 > /dev/null 2>&1; then
        echo "MQTT is UP. Let's go!"
        break
    else
        echo "MQTT is DOWN. Waiting..."
        sleep 3
    fi
done

#sh /usr/sbin/entrypoint.sh

if ! [ -f ${OTR_STORAGEDIR}/ghash/data.mdb ]; then
    ot-recorder --initialize
fi

ot-recorder --http-host ${OTR_HTTPHOST:-0.0.0.0} --http-port ${OTR_HTTPPORT:-8083} $OTR_OPTIONS ${OTR_TOPICS:-owntracks/#}

[jpmens]

That start.sh looks useful. Can you enlighten (i.e. teach) me why you switched from ENTRYPOINT to CMD?