<sanha> @ark3 Exactly. A cannot find B in DNS by B's k8s service name.
I'm trying with various languages like Node, Java in VSCode, IntellIJ... but nothing works.
A can reach out to B only with it's public full domain (like any other process or machine without Telepresence).
Please let me know the thing you are familiar with.
For environment variables, fixing source codes to pull the k8s specific env variables does not sound good.
<ark3> Issues 958 and 959 might be relevant.
<ark3> Re environment variables, it's kind of the other way around. Popular methodology suggests keeping configuration separate from code by loading configuration from application-specific environment variables. k8s lets you set env vars per deployment, so it's easy to run different configurations without building different images. See also: https://12factor.net/config
<sanha> Oh I see. I'll try with vpn-tcp.
I hoped to debug many apps with Telepresence in inject-tcp mode :(
But Telepresence is still useful for debugging single app. Thanks.
For env variables, I misunderstood the usage of them.
I guessed that it will be used for VPN or DNS of Telepresence, but it seems not.
Thanks for your explanation!
<sanha> Hi!
Unfortunately, I couldn't find any solution. It seems that the
Unfortunately, I couldn't find any solution. It seems that the
telepresence process can't locate the IDE process in my case. (My guess)
<rdl> do you have a writeup? would love to see how you got it to work (and share the blog)
<gregw> My team has adopted telepresence for local development/testing of one of our microservices running in EKS but I am using a centos-based distro of linux and am having a bad time trying to get it installed. First, I tried following the instructions for installation on Fedora-based distros, but attempting to install telepresence through dnf results in 'no match for argument: telepresence' even though I have the datawireio_telepresence yum repo config in /etc/yum.repos.d. Next, I tried installing from source and ran into another error during the telepresence build: "ImportError: No module named 'zipapp'"
<gregw> I have python 3.7.1 installed via pyenv and I've tried the install from source through a virtualenv as well. The most confusing part is if I open the python repl using the same python interepreter and execute "from zipapp import create_archive" it does not produce an importerror
<ark3> @gregw I'm interested to learn more about your error when installing from source. That should work!
<ark3> The install script runs
python3 from your shell path.
<gregw> Yeah and I have a python3 binary in my $PATH under $HOME/.pyenv/shims/python3
<gregw> ah it has to do with running sudo...if i run that command without sudo then it finds the modules and executes python code
<gregw> makes sense since the superuser doesn't have the same python environment as my logged in user
<gregw> thanks!
<ark3> Hmm. We don't test that installer with sudo. That's part of the problem here. It's only the last few steps that need privileges. Perhaps we should split up the "build binary" and "install binary" bits.
<JM> Morning; I'm wrangling a network traffic routing setup. I am on OSX and hitting a single node k8s installed in an EC2 instance. I can telepresence the deployment just fine, but I want to run a PHP debugger locally. In an "all software on the same host" setup xdebug would connect to
localhost:9000 and the IDE would have a listener present to pick it up. In a non-telepresence host-run docker container, you'd change localhost to host.docker.internal or docker.for.mac.localhost and the traffic would show up on the OSX host. I suspect that's not quite working right due to the various networking loops going on with telepresence, however. Any suggestions for how to get that connection from my --swap-deployment container to show up on my hostmachine so the IDE can pick it up? Thanks!
<ark3> I have a solution for you, but it's somewhat convoluted.
<JM> Hit me 🙂
<ark3> Telepresence sets up and exposes an ssh server in its network container, then uses
ssh -R to forward a port in the container to something on the host. This avoids using host.docker.internal, which doesn't work on Linux.
<ark3> You can use that same ssh server to set up a second port forward.
<JM> I'm open to trying that out; what would the command for that look like?
<ark3> Yup, working on it.
<ark3> You'll see a
docker run command that has --publish=127.0.0.1:port:38022/tcp
<ark3> That port, the one associated with 38022 in the container, is the port on your host machine exposing the ssh server.
<ark3> Now, to run your client, you'll want some extra options to ssh.
<ark3> Search the log file for
Launching Local SSH port forward to see what I mean.
<ark3> Does that make some amount of sense?
<JM> yeah, I think so. Let me see if I get it right; I'll be ssh'ing into the local networking container and having it do ssh tunnel proxying so that the connections from the local deployment container (since it shares the network namespace) on 9000 will pop back up on my mac
<ark3> Yes, exactly.
<JM> Bingo
<ark3> Success, with the debugger and all?
<ark3> Excellent.
<JM> I haven't quite resolved the best way to get the proper container name, but
docker inspect --format='{{(index (index .NetworkSettings.Ports "38022/tcp") 0).HostPort}}' telepresence-1554143629-148726-77228 gets you the right port to connect to