We don't want to introduce risk



  • We were a few days from a release when I was asked to help one of the younger members of the group with the release process.

    It turns out to be about 130+ easy-to-mistype commands that must always be run with the same parameters in the same order, and the process must stop if there's an error, so that it can be addressed. We use Outlook on our PC's, but X11/Xterm into Linux, and our very old crappy version of the software doesn't let you cut-n-paste across environments, which means lots of manual retyping of commands.

    Gee, I wish there was some tool that could make this task a little easier. What would help make the process automated? Hmmm...

    Oh wait, there's this 40 year old tool called make, and it's offspring: ant.

    Ok, I collect all the commands, build an ant build.xml file and test it. Then I jam it in Hudson with the appropriate targets to build and run our automated tests. The app compiled, the jars got built, the app launched and the tests ran; everything passed. All was well.

    At the team meeting, I let them know that it's set up and documented, and that we should abandon the manual process. The response (to the younger guy)?

    No, keep doing it manually because we don't want to introduce the risk of a new process this close to the release.



  • BINNED

    The old preference for the devil you know, I guess. Whoever said that may have gone through a sorcerer's apprentice type scenario with an homegrown "automation" tool that went awry.



  • I would say fine, if they can substantiate (with real numbers) that a fully automated and tested deployment method presents more "risk" than human error on a highly tedious and error prone deployment with 130+ manual steps.  Humans are terrible at this monotonous shit, by-fucking-design...

    This is a perfect time to write a "memo to note" email to all the technical stake holders, and perhaps CCing their bosses.  When shit hits the fan, probably not on the first deployment, but maybe the second or third release, you'll have your ass thoroughly covered.



  • I love how nobody else thought "If only we had something that could accurately execute a bunch of sequential instructions".



  • @frits said:

    I love how nobody else thought "If only we had something that could accurately execute a bunch of sequential instructions".

    These people haven't yet mastered the concept of independent thought...

    It's pathetic funny sad.

     



  • @snoofle said:

    No, keep doing it manually because we don't want to introduce the risk of a new process this close to the release.


    I know a guy who got a summer internship in a big University, he was tasked to do a user migration from Novel to Active Directory. It was bleeding-edge because they were doing Extreme Account Creation: one guy creating the account, the other guy double-checking it (and swapping who is typing every other day). He proposed to create a script and automate this but it was deemed too risky. So the next proposal was to have TWO scripts: one to extract and load Novel account info in a database, and the other one to pull data and create accounts in AD after the database has been checked. But it was also too dangerous. So they actually created thousands of accounts manually over the summer.



  • @thistooshallpass said:

    I know a guy who got a summer internship in a big University, he was tasked to do a user migration from Novel to Active Directory. It was bleeding-edge because they were doing Extreme Account Creation: one guy creating the account, the other guy double-checking it (and swapping who is typing every other day). He proposed to create a script and automate this but it was deemed too risky. So the next proposal was to have TWO scripts: one to extract and load Novel account info in a database, and the other one to pull data and create accounts in AD after the database has been checked. But it was also too dangerous. So they actually created thousands of accounts manually over the summer.

    Sounds like a perfect situation to write a script which extracts the Novell account info and fills in the textboxes in the AD dialog. Then both guys can check it and press OK.



  • @pjt33 said:

    @thistooshallpass said:
    I know a guy who got a summer internship in a big University, he was tasked to do a user migration from Novel to Active Directory. It was bleeding-edge because they were doing Extreme Account Creation: one guy creating the account, the other guy double-checking it (and swapping who is typing every other day). He proposed to create a script and automate this but it was deemed too risky. So the next proposal was to have TWO scripts: one to extract and load Novel account info in a database, and the other one to pull data and create accounts in AD after the database has been checked. But it was also too dangerous. So they actually created thousands of accounts manually over the summer.

    Sounds like a perfect situation to write a script which extracts the Novell account info and fills in the textboxes in the AD dialog. Then both guys can check it and press OK.

    That would be considered blasphemy (aka major non-conformity) in any ISO-9000 certified or ITIL-worshiping organization. There is no way around those big red binders. Follow the Process, or perish.



  • @C-Octothorpe said:

    Humans are terrible at this monotonous shit, by-fucking-design...

    Web Hero of the day award goes to C-Octothorpe.



  • I would've let you at least test it alongside the old process and compare outcomes. With enough sample comparisons, you can get super-duper certain that the new process does not materially differ from the old in terms of outcome.

    Sadly, the same people who don't *get* practical innovation usually don't get statistics either.

    People like that are the reason our GDP growth is flatlined.  I do not envy your workplace.


Log in to reply