We want admin to some how know that rpm -qf /etc/systemd/system/continer.service
Was created by CONTAINER-FOOBAR. Having the container create an rpm on the fly that takes into case the name of the container. On 11/18/2016 10:49 AM, Jakub Filak wrote: > Yes, you are true about the chroot. I realized it a bit late and I didn't > want to spent too much time on a proof-of-concept script. > > Could you please tell me more about the attributes you have on mind? I am > afraid that I am caught in my use case and I cannot see anything beyond that. > > > On 11/18/2016 04:24 PM, Daniel J Walsh wrote: >> I think you should be able to do this totally with a chroot /host >> >> rather then nsenter. >> >> A little trick I have been playing with for scripts executed in the chroot. >> >> Then you could just copy the rpm out of the container on to /host/run >> and then execute >> >> the ./install.sh command to execute the appropriate rpm commands on the >> host. >> >> The difficult part is creating the rpm with attributes back to the >> container. >> >> >> On 11/18/2016 09:26 AM, Jakub Filak wrote: >>> The script was initially developed on Fedora Rawhide, so it works there. >>> >>> I had to add a couple of hacks to be able to run it on Atomic. Mainly >>> because Atomic does not have rpm-build, which itself has tons of >>> dependencies, and I decided to run the script in a container. Soon I found >>> out that not only missing rpm-build is a problem but "rpm -i" does not work >>> on Atomic. However I overcome this problem with a wrapper for rpm [0]. >>> >>> The bottom line is that to make the script working on the current Atomic, >>> you must install rpm-build (possibly in a privileged Fedora container that >>> shares PID NS with the host) and issue the below command: >>> >>> $ PATH="./atomic-host:$PATH" ./af install --rpm <container_name> >>> >>> On Fedora you just need to run only: >>> $ ./af install --rpm <container_name> >>> >>> >>> Full example: >>> >>> [host] $ docker pull elcolio/etcd >>> [host] $ docker run -d --name etcd elcolio/etcd >>> [host] $ docker run --privileged --pid=host -it --rm fedora sh >>> [cntr] $ dnf install rpm-build git >>> [cntr] $ cd tmp && git clone https://github.com/jfilak/af && cd af >>> [cntr] $ PATH="./atomic-host:$PATH" ./af install --rpm etcd >>> [cntr] $ exit >>> [host] $ rpm -qf /etc/etcd.conf >>> C_etcd___docker.io_elcolio_etcd-latest.0.noarch >>> >>> For sake of simplicity, I assume that the /exports/hostfs/etc/etcd.conf file >>> exists within the container. >>> >>> >>> Jakub >>> >>> PS: The script is just a proof of concept that I created over night. >>> >>> >>> 0: https://github.com/jfilak/af/blob/master/atomic-host/rpm >>> >>> >>> >>> On 11/18/2016 01:52 PM, Daniel J Walsh wrote: >>>> Seems like a simple fix. rpm-ostree should be modified to support file >>>> path rpms as well as >>>> >>>> rpm repositories. But will this work on a traditional rpm based system >>>> like RHEL or Fedora >>>> >>>> Workstation? >>>> >>>> >>>> On 11/18/2016 03:35 AM, Jakub Filak wrote: >>>>> I've been playing with privileged containers delivering services for D-Bus >>>>> system bus. These D-Bus services must be enabled by a configuration file >>>>> placed in the /etc/dbus-1/system.d/ directory. Therefore my containers >>>>> must >>>>> install files on Atomic host and this action creates system files not >>>>> owned >>>>> by any rpm package. >>>>> >>>>> Last week I wrote a script that creates an intermediate rpm package from >>>>> files in the /exports/hostfs/ directory in a container and installs the >>>>> package to a host [0]. Unfortunately, the script uses 'ostree admin >>>>> unlock' >>>>> before running 'rpm -i ...', so my changes to rpm database disappear with >>>>> reboot. Using 'rpm-ostree pkg-add ...' could make the changes persistent >>>>> but >>>>> the command does not accept local files. >>>>> >>>>> My question is that will it be possible to install local rpm files via >>>>> 'rpm-ostree pkg-add'? >>>>> >>>>> >>>>> >>>>> Jakub >>>>> >>>>> 0: https://github.com/jfilak/af >>>>>
