public ips node filter is not working

khaledyoussef24 commented 8 months ago

Description

public ips node in load nodes is not working env used : devnet version : https://github.com/threefoldtech/tfgrid-sdk-ts/commit/5b50681261cf6e97e63a9dc4eac5535210aaae9b

AlaaElattar commented 8 months ago

Work In Progress (WIP):

after investigation. i found that checking farm ips is done after merging workloads.
still looking for a way to handle it earlier.

AlaaElattar commented 8 months ago

I tried to reproduce issue but it didn't appear.
This might have happened because there was outdated information on indexer before updating chain. (now it should have picked up the latest information)

khaledyoussef24 commented 8 months ago

can not produce the issue anymore . tried the same scenarios as last time but did not work. will re-open if i could produce it again.

khaledyoussef24 commented 8 months ago

the issue still exist telegram-cloud-photo-size-4-5875128825996754705-y

sameh-farouk commented 8 months ago

the processor update public_ip table when it create or update a contract. so this related to the the contracts issue and should be fixed when the ops deploy the new version (request)

AlaaElattar commented 8 months ago

Blocked on https://github.com/threefoldtech/tf_operations/issues/2243

AlaaElattar commented 8 months ago

As Graghql released to v2.11.6, the problem shouldn't exist anymore.

khaledyoussef24 commented 8 months ago

made a four or more deployments with public ip and did not face the same issue again

the issue is no more longer reproducible

Mahmoud-Emad commented 7 months ago

I remember we threw this error to inform the user if the selected farm has enough IPs. The question is, why doesn't it pass the IP field when filtering to get nodes?

as you see here these are the nodes I got

AlaaElattar commented 7 months ago

Work In Progress (WIP):

I deployed vm with public ip and the filtration worked fine.

Then, I tried to deploy caprover but i got no nodes were found as i took last public ip.

I deleted the vm to free the public ip, then went to deploy caprover.

when i tried to deploy worker to the caprover i got no nodes were found.

AlaaElattar commented 7 months ago

Work In Progress (WIP):

I deployed vm with public ip and the filtration worked fine.

Then, I tried to deploy caprover but i got no nodes were found as i took last public ip.

I deleted the vm to free the public ip, then went to deploy caprover.

when i tried to deploy worker to the caprover i got no nodes were found.

i was also checking the public ips on freefarm from graphql at the same time, and there was no difference at any step.

AlaaElattar commented 7 months ago

also manually the filtration works fine

AlaaElattar commented 7 months ago

I remember we threw this error to inform the user if the selected farm has enough IPs. The question is, why doesn't it pass the IP field when filtering to get nodes?

as you see here these are the nodes I got

unable to reproduce this

AhmedHanafy725 commented 7 months ago

I think what happened is a race condition. Someone was deploying at the same time and he got the last public ip in the farm before you. And nothing can be done in this case.

khaledyoussef24 commented 7 months ago

is there any solution that we can do so we can avoid this concurrency? or shall we close it for now ?

AhmedHanafy725 commented 7 months ago

This can rarely happen and TFchain already rejecting the deployment in case the ip got taken. We will support this #1822 to make the client reject the deployment too. But nothing we can do in the playground.

threefoldtech / tfgrid-sdk-ts