#django

i have been majorly stuck trying to server the static admin files in production. based on the hours ive spent on this, my configuration (gunicorn, whitenoise) seems fine

but no matter what, i cant get the static files in production

furthermore, if i enable the whitenoise STATICFILES_STORAGE, i get server error 500 responses

im deploying to heroku and it seems like the issue might be that collectstatic is failing but i have had no luck fixing it

badfilms, suggest saying what the actual error is. as in the traceback you get

my logs show 404 errors for the /admin content

ie css/responive.css etc

and all the other static files are working?

my static files are limited to admin for this project

did you follow the django documentation for setting up static files in production?

yes

you have STATIC_ROOT in your settings?

what are some good uses for django's TIME_ZONE variable? i can't think of any

i have static_root

i have whitenoise in middleware

etc

for what it is worth, this is dockerized

badfilms, does findstatic find the files?

i basically have two situations right now. if i exclude STATICFILES_STORAGE, i can visit /admin with no styles. if i include STATICFILES_STORAGE, i get server error 500 when visiting .admin

s/.admin/\/admin/g

surely if you have a 500, you'll get an exception in your logs?

also a crazy thought, is admin in your installed_apps?

yeah

i can deploy locally no problem

Oh, are you setting the storage to the manifest storage?

if i run findstatic on core.js, it returns /usr/local/lib/python3.6/site-packages/django/contrib/admin/static/admin/js/core.js

yeah whitenoise.storage.CompressedManifestStaticFilesStorage

in deployment, does the directory "staticfiles" exist?

thats where i think the problem lies

it does when i explicitly create it, but heroku doesnt seem to have it persist

it needs to exist before collecting statics

wait really?

yep

it's noted in the heroku guide

if this fixes this problem, i might kill myself for wasting 12 hours on this, but also hug you

still server error 500

dyno=web.1 connect=0ms service=13ms status=500 bytes=234 protocol=https

anything show up if you enable debugging?

heroku config:set DEBUG_COLLECTSTATIC=1

I know it's daft, but is it possible you have DISABLE_COLLECTSTATIC set too?

debug_collectstatic is 0

i'll need to make a few changes before i can ship with debbing enabled

1 sec

Hi all.

I have a potential issue.. or discussion, lets assume i have 2 models M1 and M2, where M1 has a ForeignKey to M2. Lets also assume that half of M1 objects point to one instance of M2, and the other half point to the other M2.

So 100 M1's that use 2 different M2's

If I query all M1's and use .select_related('M2'), and then iterate over M1's.. and while accessing an M1, I perform an operation on the m1.m2, and save it...

Do the rest of my M1's in the queryset have a 'cached' copy of m1 before that save?

This appears to be the case in my situation..

If I'm correct, can I avoid that, or turn that off.. would it help to remove it from select related?

do you mean cached copy or m2?

cybo_: i broke something lol... now im getting bad request 400

yes. say I do this: m1[20].m1.value = 'X' then m1[20].m1.save() then, later print(m1[21].m1.value)

Assuming m1[20] and m1[21] had a FK to the same m2

kutenai, you should keep a reference to m1. I've had cases where it refetches in the past, so the value assignment never persists

That's what I think is happening.. if I modify an m2 through one FK relationship, the other FK relationships don't appear to have that modification, and they attempt to modify this again

In my case, m2 is a "symbol" object, and it has a generated base64-image. I iterate over the main objects that use the image, and many of those use the same image.

If any of them find that the symbol image isn't set, they will calculate that, set it, and save the object, but I was seeting the same object saved repeatedly.. which should not have happened.. in theory.

is this dj2.0?

yes

oh..

sorry, no, 1.11

yeah I think this is "fixed" in dj2

ah......

I vaguely remember something going in to better share objects. however I also remember it being an internal detail. so don't rely on it perhaps

cybo_: so i can confirm that the issue is that collectstatic is not working. this is supposed to be handled by heroku automatically during deployment

badfilms, and enabling staticfiles debug doesn't show any details?

oh right i forgot about that step :p

ah, kutenai, I think what I said only applies for prefetch. in a select related, the instances aren't associated

right. its getting on for 3am. g'night

kutenai: I think that's the use case for Model.refresh_from_db()

Hmm. I'll look at that

kutenai, why not just iterate over M2 instead?

That's what Im doing now. I added a new mgmt script to update all of the m2's first.

I guess I was surprised I had to do that.. but, after some thought, realized why that might be the case.

kutenai, you could also mark locally the ids you're updating

yeah.. that would kind of break some encapsulation. Or, it would require I write a new model..

didn't want to deal with all that.

fair enough

Thanks for all of the help/advice. Much appreciated.

cybo_: DEBUG=True == no issue

the /admin static pages work fine

anyone hiring?

theology: what can you do?

where are you located?

wow... so far through this entirely API driven django app... and haven't used nap at all, yet