• Run into that type all the time. :)


  • i also meet his double from an alternate reality frequently:

    • theres a product that works and costs $50
      –-> yea but can't you combine 10 free opensource packages, hack them together, and do the same thing?

    sigh

  • LAYER 8 Netgate

    Amen, @heper.

    ETA: This is right there with wanting to bridge a quad-port i350 card instead of going to Frys and getting a $30 gig-e switch.


  • Now we just need a comic about cargocult.

    I've got a fun anecdote about cargo cult everything. Mind you, everyone that I have ever worked with in my company has been a well meaning person, but not everyone understand what they should understand to properly manage their responsibilities. We acquired another smaller company that we had been working with for a while. They had had most of their data stored in Mongo, so they had a lot of expertise in Mongo from programmers to admins. As a larger company, we have been looking at a large NoSQL database to handle our huge amounts of data which we currently have to shard per customer in SQL because we have a lot of data.

    We hired out to the enterprise consulting services of MongoDB and had them work with our newly acquired Mongo experts to create us a DB to replace a subset of our services. When all was said and done, our dual socket 10 core Xeon(20 cores totals, 40 threads), with 512GiB of ram and packed full of SSDs running on the metal is slower than quad core sql VM with 8GiB of ram using 10k-rpm SAN backed storage.

    Turns out at no point did anyone ask about how much data we planned on having or how many different ways we need to access the data in order to do real-time analytics. Mongo is crazy slow if you don't index your data, but indexes eat up memory.  Of course when I was just having a casual conversation with the DB admin, the first few questions I asked were about such issues. He told me if anyone in the room full of experts asked any of the questions that I asked in the first few minutes of talking to him, that they would not be having any of these issues. And I know nothing about Mongo.

    The cherry on top is when a senior engineer explicitly requested that the Mongo master be configured to use async replication to the slaves to maintain high write performance, then was flabbergasted that there was a race condition when he designed the code to write to the master, then read from the slaves. For performance. This race condition never occurred in dev or testing, but pathologically happened up-to 20% of the time when the system was getting hammered. Load testing, what's that?

    In the end, we now have two dual socket systems with 512GiB of ram each and about 20 SSDs in mirrored raid, and the system is as fast as cold molasses.

    It's quite frightening how people in senior positions, even under the guidance of seasoned veterans and high priced consultant services from the company that makes the product, can completely ruin a simple project with so many fundamental mistakes while following best practices.

    Cargo Cult all the way. When people don't understand what they're doing, the only way they can learn is from mistakes. Don't trust these people with spearheading new projects.


  • We run into people like that a lot….Usually IT consultants and IT Directors that see our Bill of Materials has pfSense on it.  It spurs a lot of debate and complaining until you mention that 99.99999999% of the internet runs on Linux.  Then you hear a lot of uhhhhh.....but.....uhhhhh.....Never mind.