Closed bradrydzewski closed 5 years ago
@jimsheldon I pushed a new minor version of the CLI with the new network_mode
field. I converted the yaml you provided to the 1.0 format using drone convert
, added network_mode
, and executed the pipeline and it was success 🎉
kind: pipeline
name: default
steps:
- name: integration
pull: if-not-exists
image: duffqiu/dind
+ network_mode: bridge
0.8 had an implementation flaw where it attached containers to both the user-defined network (created for each pipeline) and the bridge network. In 1.0 we fixed this and removed the bridge network. There are some pipelines that relied on the pipeline being attached to both networks, which caused issues when porting to 1.0
I believe there are few possible solutions to this issue:
network_mode: bridge
for the pipeline step (and while we are at it, we can addhost
networking andcontainer:<id>
networking as wellI am partial to option number 1 because we should expose this flag in the yaml anyway, for more comprehensive docker support. I am not a huge fan of option 2 because in some cases, attaching to both network (especially on kubernetes) can cause issues.
cc @jimsheldon this should be a simple fix either way. I will proceed with number 1 since we should support it regardless. Lets touch based Monday to make sure this is an acceptable option.