lsh-0
commented
9 years ago I may have just described this better in an email:
_it's more like building a script to generate scripts. The smaller and less they know about the world the less likely they are to failure. Something like, "./manage.py installmonitors" which would then install cron jobs like
*/5 * * * * PATH=~/.virtualenvs/mondroid/bin/:$PATH python /path/to/fetcher.py elifesciences.org > /var/log/mondroid/elifesciences.org.log
*/5 * * * * PATH=~/.virtualenvs/mondroid/bin/:$PATH python /path/to/fetcher.py ubiquitypress.com > /var/log/mondroid/ubiquitypress.com.log
ajrbyers
as an system admin running the monitor I need the logic that does the checking to fail rarely, if ever because the data being collected may be subject to auditing and intense scrutiny
I love how convenient things are once setup, but I made a dumb mistake while tweaking the configuration and didn't notice for about ten minutes when an expected state failed to appear in the list. I can't afford to have a 'droid with a faulty memory because we're building in features quickly. Rigorous unit testing might also help here before deploys, but there always seems to be edge cases with these larger, sophisticated frameworks.
This is slightly audacious, but I was thinking about having a Python script generate other, dead simple, scripts (one per monitor) whose sole purpose is to ping a website just once and output the result to a log file. This script would then be called repeatedly by cron and the log would accumulate. Mondroid then reads this log data in regularly with a management task and populates it's db.
Waddya think?