On startup, have services print their name and version info

[melange396]

On startup/restart, have services print (log) their name and some sort of versioning info, like docker image name + build date, and git repo + branch + tag + gitref (commit id), and etc...

Also print/log the current time (obviously), as well as the command line and arguments provided (though this could be problematic where passwords are provided by command line arguments)

Docker metadata isnt included in or exposed to running containers by default, so we might need to get that by adding something to our image build process (by adding stuff to the image filesystem).

The services where this should be applied include acquisition csv processing, acquisition metadata generation, webservice/api, and pipelines. We can find many of the places to insert this by looking for references to "__main__" in the code (by running grep -rl __main__ . -- ill put the output of the files found by this command in the first comment) but this is not necessarily exhaustive depending on how jobs are brought up (gunicorn, for example, doesnt seem to need main to start the webapp).

This issue is very closely related to cmu-delphi/covidcast-indicators#1702

[melange396]

Files to add this to include (but may not be limited to) :

./src/acquisition/afhsb/afhsb_csv.py
./src/acquisition/afhsb/afhsb_update.py
./src/acquisition/cdcp/cdc_dropbox_receiver.py
./src/acquisition/cdcp/cdc_extract.py
./src/acquisition/cdcp/cdc_upload.py
./src/acquisition/covidcast/delete_batch.py
./src/acquisition/covidcast/migrate_epidata_to_v4.py
./src/acquisition/covidcast_nowcast/load_sensors.py
./src/acquisition/covidcast/signal_dash_data_generator.py
./src/acquisition/covid_hosp/common/utils.py
./src/acquisition/ecdc/ecdc_db_update.py
./src/acquisition/flusurv/flusurv_update.py
./src/acquisition/fluview/fluview_notify.py
./src/acquisition/fluview/fluview_update.py
./src/acquisition/fluview/impute_missing_values.py
./src/acquisition/ght/ght_update.py
./src/acquisition/ght/google_health_trends.py
./src/acquisition/kcdc/kcdc_update.py
./src/acquisition/nidss/taiwan_nidss.py
./src/acquisition/nidss/taiwan_update.py
./src/acquisition/norostat/norostat_add_history.py
./src/acquisition/norostat/norostat_update.py
./src/acquisition/paho/paho_db_update.py
./src/acquisition/paho/paho_download.py
./src/acquisition/quidel/quidel_update.py
./src/acquisition/twtr/healthtweets.py
./src/acquisition/twtr/twitter_update.py
./src/acquisition/wiki/wiki_download.py
./src/acquisition/wiki/wiki_extract.py
./src/acquisition/wiki/wiki.py
./src/acquisition/wiki/wiki_update.py
./src/server/covidcast_issues_migration/proc_db_backups_pd.py

and especially:

./src/server/main.py
./src/acquisition/covidcast/covidcast_meta_cache_updater.py
./src/acquisition/covidcast/csv_to_database.py

[melange396]

here are some ways to get git metadata:

# repo name (and then some)
git config --get remote.origin.url

# branch and hash ref
git log -n 1 --format=format:'%D @%h [%H]'

and possibly still useful but probably less so:

# branch and remote and some meta maybe
git status -b | head -2

# a hodgepodge... (sometimes also tells you if its dirty (not reliably?))
git describe --long --tags --always --dirty

# "nearest" tag (if there is one?  sometimes?)
git describe --abbrev=0 --tags --always
git describe --contains `git rev-parse HEAD` --always

[melange396]

it may also be worth including a list of all of the installed packages/libraries and their version numbers

[melange396]

we may also want to print the values of all of the environment variables, though like with command line args as listed above, this has a chance to possibly reveal passwords.

[melange396]

it may also be worth including a list of all of the installed packages/libraries and their version numbers

This can be done with the simple pip freeze command, or with something more intricate like the procedure described at https://stackoverflow.com/a/69081814. It would be helpful if we could generate an alert when any of the versions change, as a way of cross-referencing with other events like errors or changes in performance.

[melange396]

Updated git log format suggestion:

git log -n 1 --format=format:'%h @ %ci %d'

More ideas for logging dependency environments programatically (each has its drawbacks) :

# uses nonpublic submodule  :(
# shell command rough equivalents:
#     pip3 freeze --all
#     pip3 list --include-editable --pre  --format=freeze
from pip._internal.operations.freeze import freeze
for requirement in freeze():
    print(requirement)


# deprecated, doesnt work past python 3.9ish?  :(
import pkg_resources
dists = [str(d).replace(" ","==") for d in pkg_resources.working_set]
for i in dists:
    print(i)


# not supported til python 3.10ish?  :(
from importlib import metadata
for dist in metadata.distributions():
    print(f"{dist.name}=={dist.version}")


# dont do this, it loads all the packages into memory
import pkgutil
for p in pkgutil.walk_packages(onerror=lambda e: print(f"WHOOPS: {e}")):
    print(p.name)
import sys
for module in sys.modules.values():
    if hasattr(module, "__version__"):
	print(f"{module.__name__}=={module.__version__}")
    else:
        print(module.__name__)