docker: integrate flux-accounting with fluxorama interactive image about flux-accounting HOT 3 CLOSED

the dependencies for flux-accounting need to be linked in order for the subcommands to work. My hack was to run pip3 install -r requirements.txt --user again, but maybe there is another way to do this in the Dockerfile while the image is being built?

Okay, I think I was able to solve the No module named 'pandas' error by installing the dependencies to /usr/local/lib64/python3.6/site-packages/ instead of --user, which I think was installing them to root/.local/.

Old pip3 command in Dockerfile:

$ pip3 install -r requirements.txt --user

New pip3 command in Dockerfile:

$ pip3 install -r requirements.txt -t /usr/local/lib64/python3.6/site-packages/

from flux-accounting.

Here is the final Dockerfile that I sent out to the stakeholders to interact with:

FROM grondo/fluxorama:latest

LABEL maintainer="Christopher Moussa <[email protected]>"

COPY . /usr/lib64/flux-accounting/

RUN yum -y update \
  && yum install -y python3-pip python3-devel git \
  && cd /usr/local/bin \
  && ln -s /usr/bin/python3 python \
  && pip3 install --user --upgrade pip \
  && pip3 install tox

# go into flux-accounting, install dependencies, and run unit tests
RUN cd /usr/lib64/flux-accounting/ \
  && pip3 install -r requirements.txt -t /usr/local/lib64/python3.6/site-packages/ \
  && make install \
  && make check

# load job-archive module
RUN printf "#!/bin/bash\nflux module load job-archive dbpath=/var/lib/flux/jobs.db\n" \
        >/etc/flux/rc1.d/job-archive \
  && chmod +x /etc/flux/rc1.d/job-archive

# make directory for flux-accounting database
RUN mkdir /var/lib/flux-ba

# create flux-accounting database where fluxuser is the "admin"
RUN cd /usr/lib64/flux-accounting/accounting/ \
  && ./create_db.py -p /var/lib/flux-ba/FluxAccounting.db \
  && chown -R fluxuser /var/lib/flux-ba/

# load job-archive module
RUN printf "#!/bin/bash\nflux module load job-archive dbpath=/var/lib/flux/jobs.db\n" \
        >/etc/flux/rc1.d/job-archive \
  && chmod +x /etc/flux/rc1.d/job-archive

This loads the job-archive module and create a database to store inactive jobs. When they are placed here, the database is preserved across restarts and can be accessed with SQLite queries. Flux will not launch successfully if the job-archive module is not executable.

Inside of /var/lib/, a new directory gets created called flux-ba that will hold the flux-accounting database. This database is also preserved across restarts, and can also be accessed with SQLite queries as well as with the flux account subcommands.

I think the caveat for read/write access to the SQLite databases is the same permissions should also be reflected in the directory the database file is located in, i.e. the user must also have permissions to the directory the .db file is in [1], so that's something I would need to look out for.

from flux-accounting.

I am going to close this issue since I have a working Dockerfile to create an interactive Docker container with flux-accounting and the job-archive DB installed.

from flux-accounting.