I found that, from within docker-for-mac-beta containers, the docker host could be found and connected to at the usual 172.17.0.1 address (assuming the service bound to 0.0.0.0). This comment has been minimized. You have not supply any solution to users of Docker for mac, the ip access is very demanded, because develop with docker need direct access to the container service from docker-container and from host, port forwarding not works for this situation ( the service will have different address from different place). Also the libvmnet not so fine, but it is a solution, sudo is acceptable than nothing!!! The tap device solution need no-sudo permission. At least, docker for mac should supply some hack way to let developer make their way. For example, the hyperkit start command argument can be custom by users. At now, The docker for mac is almost useless for people. : In addition this network setup would require root access which we are trying to avoid entirely in Docker for Mac (we currently have a very small root helper that we are trying to remove). On linux is not possible to run docker without root privileges and everything is fine, on OSX you want to avoid that. What is the reason behind that? So from now on docker for OSX should be like facebook app, it does not matter that core feature is not working, but it didn't require root privileges? IMO if root privilege helps resolve that issue, then fix it and In the meantime try to achieve it without root. We are currently without core feature, that we used to, that we planned whole flow based on that feature, and now we need to resolve lots of issues. Would an acceptable compromise be to consider this an advanced feature. And when advanced features are enabled they require additional privileges? The reason I want this feature returned is to keep stuff on standard ports. We have a standard naming convention for all our services. New super mario bros wii ntsc u iso. I have DNS records created for even dev which could point to a static IP on their Mac. Anime desktop mascot download free. Each container get it's own IP, so it's always port 3306 no matter which project you are working with. Each project is truly separated just like it is in qa all the way to prd. Now, when a developer wants to connect to MySQL on their local machine they have to look at a chart to see what unique port the project has. Project A is 3307, Project B is 3308 what was Project Q again? Randomizing port is also sloppy because it doesn't allow you to save connection settings and such. You still need to look it up all the time. While this might seem like a minor issue. Switching between projects very frequently is annoying, with this standardization it removes a lot of extra steps. You want to work on a different project? Git clone and run make. EVERYTHING is setup from there on out. Multiple containers, credentials, configs, etc. Single command development setup. I just caught up on the documentation at so I understand the challenges involved. Unfortunately this is a severe impediment for using docker native in our local development environment. I had hoped to migrate away from Virtualbox/Vagrant. Our issue is that we use Consul for service discovery. For local development we want to run a mix of services from docker-compose and directly from developer's IDEs on the Mac. ![]() Since services running in containers register with consul using their docker ip:port, services running outside of docker cannot integrate. Of course this is also an issue in a multi-host cluster, but that is solved by either overlay networking (or in our case, an obnoxiously hackish introspection entrypoint script that figures out the host ip:port mappings).
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |