rvm1-ansible
rvm1-ansible copied to clipboard
ruby install hangs on user input
I noticed that installing a ruby on a remote server with a user install always hangs after some time. The ruby appears to be installed, but the ansible-playbook command hangs. I figured out what is going on by adding a redirect to the install command and keeping an eye on the output file. The rvm command asks for user input and the ansible task hangs.
I see this in the rvm output:
Ruby was built without documentation, to build it run: rvm docs generate-ri
******************************************************************************
* NOTICE *
******************************************************************************
* RVM has encountered a new or modified after_install_auto_gem file in the *
* current directory, environment variables from this file will be exported *
* and therefore may influence your shell. *
* *
* Examine the contents of this file carefully to be sure the contents are *
* safe before trusting it! *
* Do you wish to trust '/home/deploy/.rvm/hooks/after_install_auto_gem'? *
* Choose v[iew] below to view the contents *
******************************************************************************
y[es], n[o], v[iew], c[ancel]>
Ideally, rvm should have an option to force install. I posted an issue on their github. I resolved this locally by editing rubies.yml in the rvm.ruby distribution and replacing
command: '{{ rvm1_rvm }} install {{ item.item }} {{ rvm1_ruby_install_flags }}'
by
shell: "bash -c 'yes | {{ rvm1_rvm }} install {{ item.item }} {{ rvm1_ruby_install_flags }} '"
It is not a nice solution, but for now it works: ansible does not hang on the install rubies task.
@thbar @sfgeorge what do you think about this one?
That's very helpful that this workaround is documented for people who hit this issue.
But I'm wary to build that in though for a few reasons, mainly because we will be blindly saying 'yes' to any other anticipating prompts that come up, from RVM, Ruby or otherwise.
Moreover, A fresh install of RVM & a Ruby should not trigger RVM to be prompting this. This is either a bug in RVM (which I am unsuccessful at reproducing) or the result of an environmental factor on the machine in question.
@ruudgrosmann So I can try again to reproduce this, would you be kind enough to provide this info?
- Version of rvm/rvm1-ansible
- The name of the Ansible Playbook "TASK" that was running when this errored.
- A tarball or zip file of
/home/deploy/.rvm
. - The .rvmrc file(s) if they exist:
-
/etc/rvmrc
-
/home/deploy/.rvmrc
-
- Operating System and version
- Some of your ansible playbook details, including:
- Tasks that relate to rvm1-ansible usage
- All variables in variable .yml files that relate to rvm1-ansible usage
@pkuczynski it is the type of temporary issues for which I usually create a fork myself, report the problem back upstream, until the upstream problem is fixed.
Completely agree with the points raised by @sfgeorge, and we can investigate though!
Let's wait if @ruudgrosmann will come up with reproducible case, otherwise we will close...
In the meantime I have upgraded my OS and as well ansible. I have tried reproducing this issue successfully though. On the install server, I removed ruby and tried reinstalling it. After the cpu load dropped, I could see from the output of rvm (on the install server) that the ruby was installed, but on the host ansible was still busy doing nothing.
The information you asked for:
- Version of rvm/rvm1-ansible
- rvm 1.29.3 (master)
- from ~/.ansible/roles/rvm.ruby/CHANGELOG.md: the first version comment was 2.1.2. Not sure if this is the proper way to determine the version though....
- The name of the Ansible Playbook "TASK" that was running when this errored.
- TASK [rvm.ruby : Install rubies_v1]
- A tarball or zip file of /home/deploy/.rvm.
- I will create one and attach it.
- The .rvmrc file(s) if they exist:
- no rvmrc found, not for the ansible-user, nor for the deployuser
- Operating System and version
- ansible host and install server: Debian GNU/Linux 11 (bullseye)
- Some of your ansible playbook details, including:
- Tasks that relate to rvm1-ansible usage
- All variables in variable .yml files that relate to rvm1-ansible usage I think the playbook itself is sufficient
hosts: bullseye
remote_user: deploy
vars:
- rubyversie: "2.4.2"
roles:
- role: rvm.ruby
become: yes
become_user: deploy
tags: ruby
rvm1_rubies: ["{{rubyversie}}", "{{jrubyversie}}"]
rvm1_gpg_key_server: "keyserver.ubuntu.com"
rvm1_install_flags: '--auto-dotfiles --user-install'
rvm1_install_path: /data/rvm
rvm1_user: deploy
rvm1_rvm_check_for_updates: False
the rvm.ruby install: tarfile created, but I cannot attach it here. I have no graphical file manager to drag and drop or copy it to the clip board. Can I upload it somewhere?
Maybe use https://wetransfer.com?
https://wetransfer.com/downloads/1054df4ce35b3715c4fed78c1b927c2420220329094400/41515ade36c22ae2982698c1da69455620220329094430/7140b0
Thank you, @ruudgrosmann! Investigating.
@ruudgrosmann, I haven't yet been able to successfully reproduce this. Could I trouble you for some additional context?
- The value of {{jrubyversie}}
- From the install server, can you share a tarball or zip of
ls -lah /data | head -n 2
? Looking just for the permissions on the /data directory. - From the install server, can you share a tarball or zip of
/data/rvm
? - Can you share the recent final text output from the ansible host at the point that it stalled? Did it include the exact same "Do you wish to trust '/home/deploy/.rvm/hooks/after_install_auto_gem'?" message as before?
- When you ssh to the install server and run
rvm list
, do you get the same Do you wish to trust . . . confirmation prompt?
If you're curious, here's the branch where I'm attempting to replicate your setup as close as possible. And here are the test results, which are currently successful.
The output from the ansible install on the installation server is the same as in the question: the do-you-wish-to-trust message The value of the jrubybversion is not that relevant for the problem, because the problem occurs before the playbook tries to install jruby, but its value is jruby-9.2.14.0 The /data directory has owner root, group ruud and file mode 775. So user deploy is not allowed to write in it. If I log in on the install server as deploy, rvm list does not come with the confirmation request, but lists the rubies without problem.
The tarfile of /data/rvm on the installation server: https://wetransfer.com/downloads/33087c8b9958189f0499d89820ed4a3820220330080110/56682745348102f1f3a18a88d4533fb220220330080127/2a1921
By reading the rvm script rvmrc_trust I found another workaround for this problem, which does not include hacking the rvm.ruby role code. That would be to add a line to the deploy user's .rvmrc
rvm_trust_rvmrcs_flag=1
in a seperate role prior to running the rmv.ruby role. Not perfect either, but visible in the playbook.