Forking environment with PostgressSQL template on hobby plan leading to all sorts of issues
At first the container just didn't deploy, but then when i tried manually removing/creating it, my "production" environment got messed up and container there got disabled.
Steps to replicate would be creating env with postgress template, deployin it and then forking it
74 Replies
Project ID:
N/AN/A
Thread has been flagged to Railway team by @Brody.
project id is applicable, please provide it
Thank you for this flag. We are actively working on it
f96e9e13-9c55-4a12-ba6e-9962cbb6d4cb
Ye, really messed things up for me. Just after I told a reddit user a certain community-driven db is up again š
right now trying to spin again from template and it for some reason creates a plugin (??) instance on deploy and throws error.
Sorry for erratic report, i'm kinda trying to patch it up asap
this is how it looks after me deleting stuff and trying to spin up again
But the weirdest thing were "Postgress" plugin-like windows appearing in original env when I tried to deploy them on forked. And yes, I guess one of my mistakes was that after I deleted postgress in forked env (because it refused to deploy), I tried to spin up a postgress plugin rather than template in forked env
yup, no deploy (no secrets exposed here, so whatevs)
This is not your fault at all. It is on us and will be fixed soon. There are no plugins involved, just a bit of a confusing situation
I'll report back when everything is fixed š
Okay, I'll stop trying to panic recreate databases in different ways and wait for feedback from you š
This should be all fixed now
Ty, I'll try thiings out. Is forking also safe now?
yes all safe. Still some rough edges in terms of UX that we are going to work on. But your databases will get created
That's great, thank you!
oh, postgress template got removed?
can't see it in New -> Templates, but k
refresh?
hmm, will try to clear cache i guess 
odd
@jr seems all is not done š¤£
That is intended. Postgres was basically deployable through 2 menus. New service -> database and New service -> templates. To keep backwards compatbility with plugins you can just use the database . Basically just add it to your project in the same way that you would before
Or just press "New" and search "Postgres"
why can i see it listed in the templates section? old cache?
The template also had a bit different setup, with persistent storage included
I understand that it was essentially setup with dockerfile and mounted dir, etc, etc, but still was convenient
They are both the same
Anyway, right now the deployment of plugin fails.
yup taking a look
are you creating the DB in the forked environment?
nope, in the original rn
it has a weird set of vars tbh
project link?
f96e9e13-9c55-4a12-ba6e-9962cbb6d4cb
the id, right?
like I HAD redis set up, but its no longer there
Yup that is separate
Tried to add another one with same result
Yup your project is in a weird state. Looking
I do still have redis in forked env, actually.
Originally I created redis as plugin and Postgres from template
yup it is all good. Are you part of priority boarding?
Yup, I am
Sorry about the delay. Fixing a few issues here. Will report back when it is all good
you should be good to go if you start from a single environment
Okay, thank you! It's night for me, so no worries - I'll get back to this in the morning š¤
Starting from single environment - you mean just no forking, right? Or no multiple envs?
just no forking
No problem, will keep it in mind, ty.
so, main env works again now, thank you very much!
When I created a new "dev" env though, I got super weird result:
the redis to the left is something long ago deleted, the rest is what is in production env. So i deleted "dev" again, tried to create "development" (all without forking) and it failed with error
š
Setting -> Enironments -> New Environment -> Create is what I used.
this is newly created env
this is production one
not urgent for me, just giving you more info
ty again for fixing it for the main env!
i'm leaving the
test_creating_new_env be , just in case it helps you with anything, now sleep time š
thanks for all the info!
deleting things in that new env deleted them in production env too š
oh, maybe that's actually intended š
what flow did you use to delete the service?
Im dum in the morning
right click -> delete service -> type name to confirm
But now that I think about it, it kinda makes sense, to a degree, since the point of env should be same services but different env vars. But im not sure
right click -> delete service does the same thing as open service -> settings -> danger -> Delete Service from All Environments
except the modal when you right click delete doesnt give you two warnings about how it will delete the service from all environments
the action is intentional, its just missing the same warnings as the delete service from its settings hasoh, I see. Is it possible to delete from single env? or not intended for enivronments to work that way?
you want forked environments for that, but thats broken still i assume
@Angelo @Melissa another right click delete victim š¦
I edned up recreating it in new project
2c9f2913-06da-4053-a6c3-14bd48bb2682, works good so far after manually setting env vars. My only problem is that while redis spins up in development env, I can't quickly observe data tab in ui in it š¤ might've messed sth with env vars, I guess, but i certainly can curl it from outside
did you change any of the variables it came with?
well, since i copied from production initially, ye š
my dev env looks like this:
So i did change <actual password here> and <actual user here> compared to prod
i tried changing the port too, but then switched it back - i assume its okay to keep it the same, since its different ip/domain on internal network
nino
oohp, didn't edit out 
I don't even think you can specify a username other than default
lemme try to connect with redis-cli
yeah changing the username in the variables doesn't do anything, it will accept any username
i succesfully used AUTH with the password I manually set there
so it definitely works for cli
yeah just the default deploy accepts any username
True. Anyway, I can connect with redis-cli, but the ui doesn't display anything. then again, right now it's empty, hmm. I'll try ppulating and see
maybe the UI doesn't like the format of the password you set?
true, my django project doesn't seem to like it too š
eugh, lemme just reuse the one from prod
no harm i guess
and see how it goes
ye, true
will need to look up docs
possibly it's stricter format than i expected
now it displays properly
if that works, for future reference generate a password from a lowercase alphabet, uppercase alphabet, and 0-9 digits, no specials
oh, I see
the "!" got me. Was totally ok with cli though š
jr has told me they plan to add back an easy way to regen the password for databases, since currently (as you know) you have to manually make a password yourself
ye, that would be good š
bumped that to the team just now too
This is good feedback, thank you