322ea2ee-d764-421d-9d11-77dde0980f62

This is what the observability page looks like for today

322ea2ee-d764-421d-9d11-77dde0980f62

It looks like mongodb is taking a giant chunk of this?

RAM is probably the most expensiveish resource on Railway (and most platforms), your application seems to use quite a bit -between 500mb and 1gb (specially mongo)- which is billed by the minute.

My stupid dependencies… is Mongo not one of your most popular databases? Is this really just hope much it uses?

Yeah, mongo is pretty popular

So nobody can sustain a mongo database under the $5 plan, lol?

Prob PostgreSQL, MySQL and Mongo

Seems to be using 250 and rising, consistently

You can enable serverless to scale it down to zero and then scale back up based on traffic

But yeah, mongo is know for being very mem intensive, together with mysql

I don't really see that being feasable with the amount of time a coldstart would take

Feels weird that nobody can host a very popular database on one of the two plans, especially because it's like, the entire point of this system

I always recommend postgresql if the user wants lower costs

Too late for that 😂

I mean, that's mongodbs fault

I suppose ¯_(ツ)_/¯

250mb is a lot

MongoDB took ~20 cents in one day, * 30 meaning it would be $6 for just one database

@Loudbook this is what i mean (from <#727689277219012669>)

7 cents of ram and 11 cents of cpu

thats from your screnshot

oh I am fairly blind aren’t I

do you run it in a metal region by any chance? when i moved mine to metal the cpu costs went wayyy down

here's the latest today

i'll check

I am not

I’m going to migrate to a closer data center that’s also on metal

yeah definitely give that a try

mongo seems to run better on metal cpus

Is all of this amd64?

yep

Why not arm?

That would be dirt cheap

idk ask railway

lol

🚂

but anyway you can see here on your observability graph, thats an insane amount of cpu for a small db

mine did nonsense like this too before i switched it to metal, i couldnt tell you why it behaved that way but metal cpu good

it now uses so little cpu it barely even registers on the graphs

damn

I’m using 0 now

and a fraction of the RAM too

the memory will creep back up as its used

nvm still most of the RAM as before

yeah

i added --wiredTigerEngineConfigString="cache_size=50M" to mine which stopped it from going above 300 randomly, but it still hovers around 300 and theres not really anything you can do about that

You can use a free Mongo cluster if you don't host it in railway

Well yeah, but I don't want every environment connecting to the same one

As long as you store less than 500mb iirc

i switched to using one on railway when mongo killed my free cluster randomly 🥲

uh oh

had a good run tho, that cluster was alive for years

thats strange

nevermind we vibin

is your app using the db a lot?

bipolar db

nope

Always use a Mongo dump every time life feels like it's going a bit too smoothly

yeah!

i regularly do that now

Railway Pro plan gives you auto backups

It's p fire

railway pro is also

$20/m

i'll keep that in mind if my usage ever grows to $20 naturally

but i'm not paying just for the features

i fear

That's down from 110 last month

what are you hosting

Few bots and a site

do you not get constant random hangs, errors, and invulnerable volumes

No

0 issues

and that costs you $110/m?

24/7 uptime too

No there was a specific bot that was 60 alone that I no longer host

what the heck

that's a ton of money

Big bot in a big server

Lots and lots of resources

you could get a full VPS with a decent CPU for $50/m

and host all of that

Too much effort railway ux is better

Railway makes life so easy that it's worth paying the premium

But my bots cost a lot because they are high usage

And also maybe not the most efficient

💀💀

meanwhile I can't even deploy a template

due to

"unknown error"

What template are you trying to deploy

I always deploy from GitHub

tailscale subnet router

why 😭

:(

idk why they sobbed but looking at my bio may give a hint of why i sobbed

unknown error is weird tho

but maybe worth a separate thread if it persists

deployed that template now and it seems fine for me

Real

you've likely hit the service count limit on the trial plan

Ah well

If I can get my mongodb usage down I’ll consider buying hobby

as a few people have said, that's unfortunately just the reality, mongo uses a lot of memory in comparison to Postgres

Yep I'm not upset

!s