HTTPError when using a Repository Group

[tstoneba]

When using the Repository Group URL I get network errors. I can work around this by listing the repositories in the group individually as gem sources but would prefer to us a single URL. Using Nexus OSS 2.7.1-01

Error: Fetching gem metadata from http://xlddevtools01:8081/nexus/content/groups/GemsGroup/.....Retrying dependency api due to error (2/3): Bundler::HTTPError Network error while fetching http://xlddevtools01:8081/nexus/content/groups/GemsGroup/api/v1/dependencies?gems=facets%20%3E=2.0.0%0A,,language,facets_core,facets_more,daemons,eventmachine-le,http_parser.rb,preforker,shoulda-context,shoulda-matchers,faraday-middleware,httpauth,tins,mongrel,tilt,backports,rack-protection,SexpProcessor Retrying dependency api due to error (3/3): Bundler::HTTPError Network error while fetching http://xlddevtools01:8081/nexus/content/groups/GemsGroup/api/v1/dependencies?gems=facets%20%3E=2.0.0%0A,,language,facets_core,facets_more,daemons,eventmachine-le,http_parser.rb,preforker,shoulda-context,shoulda-matchers,faraday-middleware,httpauth,tins,mongrel,tilt,backports,rack-protection,SexpProcessor

[mkristian]

are you saying referencing the proxy and hosted directly in your Gemfile works without the errors ?

[tstoneba]

Yes. That works without errors.

Tony Stonebarger — Reply to this email directly or view it on GitHubhttps://github.com/sonatype/nexus-ruby-support/issues/50#issuecomment-36102888.

[mkristian]

I tried to reproduce you problem - well I did not succeed.

could you first use the latest code of the nexus-ruby-plugin and then it would be nice to see the server log (preferable with debug logging enabled) to see what is going on the server side of things.

• christian

[gmanfunky]

I think I ran into something like this too. It's like it gets into a loop or a deadlock when using the /dependencies?gems= API

This may not be called by "gem install" but it seems our chef cookbooks try to query it.

Problem behavior demonstrated with the following: curl -v http://localhost:8081/nexus/content/repositories/rubygems/api/v1/dependencies?gems=ruby-mysql

Nexus debug logs seem to show a bit of a loop and halt - https://gist.github.com/gmanfunky/9637121

I don't yet have an environment to build and try the latest and greatest, but it showed up in a snapshot pulled down January 20, 2014 - https://github.com/sonatype/nexus-ruby-support/commit/a2cd94346b45d378e50632b8c784c139415f7f65

(Also tested on a Nexus 2.7.1-01 install. )

[mkristian]

@gmanfunky so you are saying the debug log stop at the end of your gist. there were some deadlocks recently which happens when you forget to close a stream then further read-write-locks hang, i.e. I found a few places where the stream was not closed. if you can not compile the latest/greatest code then just grab the bundle from here: https://oss.sonatype.org/content/repositories/central_bundles-1850/org/sonatype/nexus/plugins/nexus-ruby-plugin/1.2.0/nexus-ruby-plugin-1.2.0-bundle.zip

actually it would be great if you could test that because in case it still shows your error, it is still possible to stop the release to maven central ;) but from my side I would say those deadlocks are gone.

currently the bundler API for the proxy uses the same code as for hosted repos and there I need all the *.gemspec.rz file from a given gem and that can take a lot of time to download from rubygems.org (even with fast internet access). currently I am working on that.

[mkristian]

from nexus-ruby-plugin-1.3.x onward the whole bundler API got a new implementation. these error should be gone unless you have a slow internet connection where some timeout on bundler will be triggered and bundler uses the specs.4.8.gz data to resolve transitive dependencies.

I consider the issue to be closed for those newer plugins.