search

ipfs-inactive / dev-team-enablement

[ARCHIVED] Dev Team Enablement Working Group
3 stars 1 forks source link

IPFS nodes on master and workers are unable to send data between them #136

Closed alanshaw closed 6 years ago

alanshaw commented 6 years ago

Which Project is this about?

ipfs/blog

Provide a link to a example run where it went wrong

https://ci.ipfs.team/blue/organizations/jenkins/IPFS%2Fblog/detail/PR-171/10/pipeline

Which platform did the issue happen on?

n/a

victorb commented 6 years ago

The issue here is that the go-ipfs nodes running on the master and worker seemingly can't send data in-between them, although they are connected (confirmed via ipfs swarm peers) and they can see each others content (confirmed via ipfs dht findprovs listing the other node). Even though everything looks correct, when running ipfs pin add or ipfs refs -r, the nodes are unable to fetch even one block from the other...

Currently they are running a old version of go-ipfs (v0.4.13) so I'm gonna try to upgrade them and see if it makes it better.

Other than that, @ipfs/go-team I'd love to have some ideas on how to debug this to know exactly where the issue is, as I cannot figure out where/what the exact problem is.

victorb commented 6 years ago

Pretty sure I found out the reason. All the linux workers are using the same Peer Identity, so Jenkins master thinks it's connected to the right peer (right ID, wrong IP). Can't believe I didn't see this before...