We get random errors in specs, that relate to user table

Good day (not actually that good for me),

I don’t know that to do here. I have one of our rails based projects going ape-shit in a random places - i basically for 2 months waste 1/4 of my time to deal with this . We’re very dependant on CI - it deploys, tests, help with code review - and it really slows our development process down for last 2 months.

Those issues happen randomly in different specs - but common thing is they all involve our most popular table ‘user’. It seems like user table get’s emptied at some point unexpectedly or FactoryGirl is not properly executed.

Non of those errors pop in locally for anyone in development team. I did 20 runs of full suite - not a single failure. I even bought a test package on travis ci - and run 20 times there - no errors.

  • I rewrote some of specs - to avoid FactoryGirl and build record with plain ruby.
  • I avoided using let() from rspec, since they are lazy evaluated
  • I tuned down parallelism feature on circleci, it lead to less errors - but didn’t solve them. (if i use 2 or 3 - then 1 build out of 5 will be successful)

I’ve submitted support tickets, two already. Pinged circleci through twitter - zero help, it seems that they don’t even read my message. Just last sentence.

This is my last cry for help - i can move to travis (i’ve tested, it at least works stably for us). But amount of work to get everything integrated - scares me and i still hope that we can resolve this.

Did anyone had same problems with CircleCi? What to do? I lost any hope to get proper support.

I basically have to click ‘rebuild’ for about an hour to get shit deployed.