then Composer nukes the WP core files & the phrase SINGLE POINT OF FAILURE echoes through your stupid head b/c everything is symlinked there
— Lisa Linn Allen (@LisaLinnAllen) December 14, 2015
I guess I would rather find the problems quicker than slower. With the symlinked WordPress core, I found a problem very quickly. Another dev had an old copy of composer.json and ran composer update in her playpen. Composer helpfully removed the copy of core that everything was symlinked to and updated WordPress directly in her playpen, I think. Something happened, WordPress was gone, everything was, as we say, hosed.
Now I make sure that the composer.json that gets checked into git doesn’t point to the directory where the current WordPress core is stored. I point it to a directory called /latest/ where it can happily update or delete or whatever it thinks it wants to do.
I still don’t like the single point of failure aspect of this, but I don’t have the UNIX chops to figure out how to make a failover should the current source of core disappear. I think a failover is what is really needed for good redundancy in the system, so it’s something I will probably pursue, but over time.
In the meantime we are working toward a deadline and we lost our project manager, so I am preoccupied with running the project and making sure the underpinnings don’t fall to pieces under us as we barrel forward. Mostly right now I worry about RAMP.