63 Replies
GitHub
Dokploy BE dies after a few deploys · Issue #455 · Dokploy/dokploy
To Reproduce Idk run my deploys i guess Current vs. Expected behavior no crash Provide environment information last ubuntu on hetzner Which area(s) are affected? (Select all that apply) Application...
out of disk space
I’m going to say that 😅
how can it run out of disk space after a few hours? looks like there needs to be an option to immediately prune old deploys
especially failed images
it consumed 40GB of space in a few hours
You are doing so many deployments?
i was testing runs yea
Got it, you can activate the docker cleanup in the settings
But that will run in everyday
yeah thats what i thought
If you are using the API you can hit the endpoint to cleanup
Manually
i have that active but it didnt get the chance to prune
well i guess for a normal run scenario it's a non issue, but for debugs yea maybe mention in the docs that it goes out disk pronto
ouch lol
You can do it trough the UI btw
You have different options, clean unused images, unused containers, unused volumes clean builder & system prune
looks like i messed it up, any idea how to reset it?
You cannot access to the UI?
nope ded
Ok in that cases you can do this
or you can do this docker service ls
docker service scale postgres=1
weird thing is its running
ok do this, docker service scale dokploy=0
then docker service scale dokploy=1
yep that did it
nothing works?
for me i can access http://188.245.149.114:3000/
it started normally and immediately started re-running last builds
Yes, this is a mechanism of the queues and bullmq that will retry if there was a server error, I think it would be ideal to remove this.
yep should just start asap to avoid downtime i guess
btw is there a way to scale to 3-4 instances without setting up container registry? just load balance between internal processes instead of multiple VPS
node is single threaded i have some heavy ops that would lock up and i'm too lazy to make queues atm
You can create multiple instances but in the same server
Without setting up a registry
In advanced -> replicas
ui is a bit confusing then
We can improve for sure
Did it worked ?
nop
🤔
What did you set?
4
Let me check
Got it, there is a wrong validation in the frontend side
error looks like drizzle foreign key constraint to me
nop, is just because we have a wrong condition here
we are passing an empty registry id ""
so that's why it throws
ahhh fml gotcha ok i will have to wait for a fix then
I pushed a new release in v0.8.2
awesome, i just managed to secure my install so all ports are locked by default instead of having 3000 etc open 😄
will try to update via interface see what happens
app tried to redeploy a working build and bricked itself, interesting
github provider died
huh
external connections should work fine weird
ok enabling ufw no good, any idea how to close 3000?
You closed the port 3000?
But you can access to the domain?
yeh i wanted to do this:
so that ufw blocks incoming connections to anything except 80 and 443
everyhthing works except github connection somehow stopped working
What happen if you try to use git provider?
sends to github ok but then redirect hangs
I don’t know if the octokit library we use have some issues internally
yeah it timed out, really weird because it should be able to connect fine
disabled uwf -> fine
Hmmm can you try the git provider?
Also I have a question you setup the GitHub account when you was in ip:3000?
Or when you was in the domain ?
nah everything immediately via domain
Strange, had no idea
first thing after creating acc -> add domain
Probably something internal of docker
yeah i didnt like that it was open but disabling iptables breaks something internally
regular fetch, idk https://github.com/octokit/request.js/blob/main/src/fetch-wrapper.ts
GitHub
request.js/src/fetch-wrapper.ts at main · octokit/request.js
Send parameterized requests to GitHub’s APIs with sensible defaults in browsers and Node - octokit/request.js
i would suggest terminate access by ip:3000 if domain is configured
I don’t do that, because probably someone can touch the traefik config and break the traefik instance
In that case you will lock the instance and you will cant access
hm can still recover via ssh no?
worked btw!
Yep, but you probably need to tweak the traefik files and maybe running some docker commands
For anyone wondering: https://stackoverflow.com/questions/30383845/what-is-the-best-practice-of-docker-ufw-under-ubuntu
this solution worked
Stack Overflow
What is the best practice of docker + ufw under Ubuntu
I just tried out Docker. It is awesome but seems not work nicely with ufw. By default, docker will manipulate the iptables a little bit. The outcome is not a bug but not what I expected.
For more d...
so you blocked everything except the UI trough the domain?
yes, though not sure if a bug but github app was created like this
i reconfigured ufw so it needs to allow external connections to docker, otherwise docker has acces to the internet but not other way around
and back online, now works
i can now confirm the app was created without domain, with domain works perfect