Having issues installing on RHEL 7 #1252
Assignees
Labels
Comments
|
You've downloaded |
Done. dcos-docker doctor then blows up looking for some azure stuff, so I had to get that all installed (kind of annoying because it blows up without a version, then when I do a pip install, it actually tells me the right version, so I have do an install again). Now it complains when it tries to build the image: pip freeze output: |
|
Hey @openshiftninja, I think I was jumping ahead to quickly. I just had a try on a RHEL 7.3 AWS instance and successfully launched a cluster following these steps (no other prerequisites required): |
|
Thank you @openshiftninja for the detailed report. Thank you @timaa2k for detailing exactly what to do!
However, we should still get the Linuxbrew issues fixed. Your original tweet mentioned a Could you please:
|
|
I got blocked on other stuff today, but I will be trying out the stuff that you guys suggested Tuesday. |
|
I did a reinstall of brew, and now trying to install dcos-e2e with it results in python3 just seg faulting: I'm not really getting anywhere with brew, so I'm bailing on that. Tried with the pip3 install + the git url and still getting the same error: |
|
Having said that, let me try after installing some of the python and other dependencies you mentioned from the sources you specified in the instructions. |
|
No dice. I installed fresh python3 and pip and created a new virtualenv from the above steps: Successfully installed all the DCOS-E2E packages: Still getting that same error when running dcos-docker doctor: |
|
Let's work on getting some more information about what that Docker build error is. |
? |
|
@openshiftninja I think that is an older version: |
|
Updated. That doesn't appear to give any additional info. |
|
Ok again thank you for your perseverance. I will look into how to get more information from Docker about the build failure. |
|
I have added a Please can you upgrade DC/OS E2E and then run the doctor command: and then paste the output. |
|
Ok, that's giving much better info. Looks to be a proxy issue, as suspected: |
|
so essentially I need to get my proxy configuration inside so that it can properly build that image. |
|
Can you take a look at https://docs.docker.com/network/proxy/ perhaps? |
|
No, this is a proxy issue for yum, not for the docker daemon. I just need to set the env vars http_proxy/https_proxy by adding ENV commands at the top: I had to do this in a couple of the Dockerfiles and then also set the -k flag on curl (unfortunately, our proxy sends back its own certificates so the verification steps fail). I've gotten further and am now hitting this error: |
|
Cool, we're getting there! |
|
|
Of course you need to specify the artifact as usual :) |
|
I'm not sure what this dcos_generate_config.sh script is... ? |
|
Oh ok. That is the DC/OS installation artifact. I was jumping ahead because usually you should do |
|
I assume we are talking about this: https://downloads.dcos.io/dcos/stable/dcos_generate_config.sh |
|
Exactly! Please try the |
|
ugh. dcos-docker download-artifact blows up due to certificate verification errors due to the aforementioned proxy sending its own certificate. Downloading the artifact now via wget. |
|
Doesn't like my wget download. Grabbing through my browser instead. |
|
Yeah this seems like a half downloaded artifact file. |
|
to be continued Wednesday. :) |
|
This looks promising: |
|
Yes it does! If that works let's still know that for you:
Let us know how the usage goes and then we can get back to the above. |
|
I did have to hack the proxy in to the Dockerfiles that dcos-docker is using, and linuxbrew still broken for me. It appears I need to configure dc-os with the proxies now, because I can't authenticate to log in (must be connected to the internet). Plus dcos-docker doctor is now spouting some new errors: Once I configure the proxy, the bottom error should go away, but the others I'm not sure about. |
I will make an issue for getting an interface for this one.
I will make an issue for narrowing this down and may ask for data from you.
Ok, let's get that done and then we can see where we are. Thank you! |
|
I'm booked on finishing something up for work this week, but I'll try to squeeze in a fix Friday and see if it makes DC/OS work. Then we can continue on the other stuff. |
|
So I put an environment.proxy file under /usr/lib/dcos using the Dockerfile that builds the base image, and it shows up in the three containers when I launch the cluster (master, slave, public agent). I can actually validate that it can reach outside because when I try to authenticate with Microsoft, it actually prompts me with a number that I validate with my Authenticator app. Still doesn't let me log in though. Pops up a generic "Unable to login to your DC/OS cluster. ..." Not sure where to look in the containers to find more detailed error messages, but looking now. |
|
hmm... tailing everything that has a .log extension in all three containers (except for the replication logs) and not seeing any messages that are giving me any hints. |
|
I do see, however, that there is a request to login?_timestamp=xxxx that has a 500 rc. There should be a log somewhere to tell me why... :) |
|
The only thing I see is some errors in the crash.log on the master: |
|
@openshiftninja Thank you for looking into this. What is the exact error message you get in the UI? |
|
Just unable to login.
|
|
So this seems to be a DNS issue. dcos-docker configures DC/OS by default with the Google DNS server 8.8.8.8 which may you be restricted from reaching by you environment/provider. Please refer to the steps described in this other open issue to troubleshoot the symptoms: |
|
Interesting. I tried github and Microsoft (as above, I actually was able to get a prompt for matching a number to my Authenticator app, opened my phone, hit the right number, but still not able to log in). |
|
The OAuth authentication flow is happening locally in your browser with your local connection and DNS resolver from your laptop. However when you're authenticated through let's say Microsoft and want to interact with the DC/OS cluster (e.g. logging in) the containers must be connected properly to the internet and be able to reach the configured DNS servers. |
|
ok, if I disable oauth authentication, I'm able to see my cluster. I was even able to get the dcos command, but the dcos cluster setup command fails thinking I need to authenticate: I tried adding a resolver and turning on oauth, but I basically then can't log in. |
|
Ok, at this point, I'm really digging in the weeds. We have a DC/OS lab installed, and I can hook my dcos command line tool up to it, so I don't see any reason to keep hammering trying to make this work. I think there are too many environment-specific issues for me to keep you guys occupied with it. I'll open another issue if I run into issues working with that lab environment. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I'll preface this with the fact that I'm aware that RHEL isn't explicitly noted as being supported, but I figured that I should be able to make it work since it's still Linux and Python.
I installed Linuxbrew as instructed on https://linuxbrew.sh:
Installed the group install of "Development Tools". Did brew install gcc (Linuxbrew suggested these in the next steps). I ran brew doctor and fixed a few things (extra -config scripts detected, git version too old, etc).
Then kicked off the install of dcos-e2e:
Ultimately, however, things are breaking with segmentation faults, namely python3:
Ok, so I am bailing on linuxbrew for now. So instead, I tried setting up a virtualenv for python:
Now I clone the git repo:
And then do a python setup install inside of the dcos-e2e folder:
Looks great, but then when I run dcos-docker doctor, it fails when it tries to build the image:
I did a pip install of azure.common in the virtualenv, but it still is giving me the same error.
The text was updated successfully, but these errors were encountered: