This message was deleted.
# atlantis-community
s
This message was deleted.
b
Still happens on Atlantis
0.24.4
with Terraform
1.5.0
j
@Bruno Schaatsbergen I just posted a workaround in the issue comments that may help you.
I just implemented the workaround this week and it has been working well for us in a repo that typically has ~30 configurations to plan in parallel.
b
Thanks Jason, I'll check it out! I simply have 2 plans, not executed in parallel. But I do have 2 separate repositories triggering A workflow at the same(ish) time causing it.. Which is a very minimal setup... 🤔
j
It WILL fetch the providers per-configuration and it does not re-use them from a common cache dir.
b
Thanks for sharing, I'll try it out
But does this cause more egress to happen, as it fetches the modules and providers everytime again?
j
Yes you are correct
It trades egress/bandwidth consumption for stability
b
Hmm ok, I'm a bit surprised to run into this issues with just running 2 workflows in parallel from 2 different repositories... I never had this happen before
j
Hashicorp need to make the cache dir functionality concurrency safe. That would be the "real" way to fix this behavior.
They use the same
~/.atlantis/terraform.d/plugins
dir if you don't explicitly change that behavior
So if both projects were running
init
at the same time you would run into concurrency problems accessing the files on disk.
My workaround just splits the cache dirs per-init instead of using one for all
b
Check, thanks for the explanation. Good to have the workaround available. Odd that I never ran into this before 🤯
j
Did you recently upgrade to terraform >1.3.x?
This started with terraform 1.4
b
I think that's the problem, today was my first client rolling out Atlantis with 1.3.8+ .... 1.5.0 actually
Dang, good to know this. Again, thanks for the fast and detailed reply Jason.
j
I am glad all of my pain could help someone else 🙂
b
😄
w
https://atlantis-community.slack.com/archives/C5MGGAV0C/p1693418545357719 Assume you saw this already, but definitely look at this if you’re seeing issues going from <= 1.3 to >= 1.4