Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Already on GitHub? Sign in to your account

Running docker on RHEL #172

Closed
davidreuss opened this Issue Mar 26, 2013 · 140 comments

Comments

Projects
None yet

This issue is meant for tracking dockers status on RHEL (Red Hat Enterprise Linux)

The one main pain-point i believe is AUFS, which isn't directly available on RHEL.

I haven't myself had as much time to play around with it, but i wanted a central location for storing progress/feedback from people trying to get docker running.

Collaborator

shykes commented Mar 26, 2013

Contributor

jpetazzo commented Mar 27, 2013

You might want to try this kernel:
http://get.docker.io/kernels/kernel-3.2.40_grsec_dotcloud-4.x86_64.rpm

Warning: the package is 700MB! I'm sorry, that's what I got when running make rpm on our kernel tree.

This kernel features:

  • grsec (for hardened security)
  • AUFS
  • setns support

I don't have any RHEL/Fedora/CentOS machine around, so I can't test it, but it's compiled with exactly the same .config as our kernels. It is configured to run on bare metal, KVM virtual machines, Xen virtual machines (and therefore supports EC2). It might work on HyperV but I never tested it there.

Collaborator

shykes commented Mar 27, 2013

cc @brianm

Collaborator

shykes commented Mar 27, 2013

Anyone else interested in Red Hat support?

bfulton commented Mar 27, 2013

+1

I'm interested in building a proper AUFS RPM against the existing sources, honestly I think having it released to EPEL or something would be best. Future maintanability and development ecosystem and so on.

I'll start researching what this effort will take and open a case and link it to this one.

Contributor

mrallen1 commented Mar 28, 2013

+1

Sent from my iPhone

On Mar 27, 2013, at 5:51 PM, Solomon Hykes notifications@github.com wrote:

Anyone else interested in Red Hat support?


Reply to this email directly or view it on GitHub.

Ideally a DKMS RPM should be built. I am pretty sure RHEL 6 uses DKMS and it will keep AUFS working across kernel upgrades.

Contributor

jpetazzo commented Mar 28, 2013

I could be wrong, but it seems that AUFS doesn't support building through
DKMS.
Unless there is a generic way to build any kind of module with DKMS?
I would be quite surprised though; because AUFS hooks itself within some
very specific places within the VFS layer.

On Wed, Mar 27, 2013 at 11:51 PM, Joseph Glanville <notifications@github.com

wrote:

Ideally a DKMS RPM should be build. I am pretty sure RHEL 6 uses DKMS and
it will keep AUFS working across kernel upgrades.


Reply to this email directly or view it on GitHubhttps://github.com/dotcloud/docker/issues/172#issuecomment-15571153
.

DKMS can build any module. It's the same as doing make, make install in the module source directory. (or doing make modules; make modules_install in the kernel source tree).

The only thing that DKMS does for you is automatically rebuild the module against the newly installed kernel so that it has the correct symbols.

Contributor

jpetazzo commented Mar 28, 2013

It's worth trying, but even when it's built as a module, AUFS patches the
VFS layer, AFAIR.

Ahh yes you are correct, my mistake. :(

Shame about that, I hope one day AUFS becomes mainline so this doesn't happen sigh.

+1 for RHEL/CentOS

It appears that DKMS will not work for this, as we cannot logically separate the components of AUFS from the Kernel image 100%. It still requires the Kernel image be patched with specific AUFS code.

The only realistic way to include this is to patch it directly into the EL6 upstream Kernel and release that code as buildable. Which is what I'm going to pursue right now. Once it's complete I'll be releasing it on my Github, however a longer term goal may be to operate a specific Kernel for Docker anyway. Include all the new fancy cgroup utils and other filesystems.

Thoughts?

Contributor

jpetazzo commented Mar 29, 2013

If you can point me to the EL6 source tree, I can add it to our kernel build farm (with some luck, our AUFS patch queue will apply neatly; or at least, require only minor work to merge).

FWIW, I recompiled our production kernels without debugging symbols (so now the RPM is just 50MB instead of 700MB):

http://get.docker.io/kernels/kernel-3.2.40_grsec_dotcloud-4.x86_64.rpm

If someone could give it a try, that would be awesome.

Depends, are you using AUFS 3.x? If so the RHEL Kernel won't work, since it's a 2.6.32 Kernel. It'll be missing most of the API/ABI that I'm assuming AUFS 3 requires since it specifically states Kernel 3.0 requirement.

However though, as requested, the current SRPM:
http://ftp.redhat.com/pub/redhat/linux/enterprise/6Server/en/os/SRPMS/kernel-2.6.32-358.2.1.el6.src.rpm

Obviously you can unpack the source using rpm, in the case you're on debian or osx (with brew) or something...

rpm2cpio kernel-blablabla.rpm | cpio -idmv

Also, going to give that Kernel a try now on an EL6 machine.

Yea, there's a lot of depsolving and conflicts in this Kernel you have posted...

Where is the srpm at? I'll be happy to work the spec out.

Ideally, this Kernel wouldn't upgrade or replace the existing RHEL Kernel (so it'll remain updated) it should be supplemental to the RHEL Kernel and updated separately.

Errors here:
https://gist.github.com/stevencrothers/5269833/raw/a42f40f6feb2813fb6a78b4e80cfcb08c1cc525e/gistfile1.txt

I'm going to open a case for this so it can be properly tracked.

For the record, new case is here to track the specific Kernel changes:
dotcloud/docker#253

This case should be made dependent on 253, not sure if that's possible on Github.

stahnma commented May 2, 2013

It seems like targeting Fedora 18/19 (and thus what will likely become EL7) would be easier/better than working on the 2.6 kernels in EL6.

Collaborator

shykes commented May 2, 2013

FYI Ubuntu is discontinuing support for aufs in their 3.9 kernels, so we
might be forced to extend support to another filesystem sooner rather than
later. I started playing with btrfs, and will continue to tinker during my
upcoming vacation :)

On Wed, May 1, 2013 at 10:47 PM, Michael Stahnke
notifications@github.comwrote:

It seems like targeting Fedora 18/19 (and thus what will likely become
EL7) would be easier/better than working on the 2.6 kernels in EL6.


Reply to this email directly or view it on GitHubhttps://github.com/dotcloud/docker/issues/172#issuecomment-17321530
.

pdf commented May 15, 2013

OverlayFS is attempting (again) acceptance for merge into mainline 3.10, but it still seems unlikely at this point, since Al Viro (the VFS maintainer) doesn't like some chunks of the design (primarily use of xattrs for opaques), which seems to be causing a little friction since Miklos is somewhat attached to the idea for portability (FWIW, I kind of agree with Miklos here).

That said, it's already included in the standard kernel patchsets for at least Ubuntu and SuSE, maybe others. and as @shykes points out, aufs is soon to be dropped for Ubuntu so some decision here will have to be made.

Contributor

jpetazzo commented May 15, 2013

I'm actively working on btrfs support for docker [1].
It's not ready yet, but it should offer a pretty solid alternative for
systems without aufs.

[1] jpetazzo/docker@master...btrfs

I just hacked support for overlayfs, which is is already on the SUSE and Ubuntu kernels, and probably the only one that will make it to upstream. I can't get the tests to pass because an unmount fails in the test. However I can run the container and docker diff seems to be working.
https://github.com/dmacvicar/docker/tree/feature_overlayfs

//cc @flavio @silviomoioli

Contributor

kstaken commented May 27, 2013

@dmacvicar Did you change FILESYSTEM_MAX_STACK_DEPTH in your kernel to allow overlayfs to stack more than 2 layers?

edit: just noticed this limit may not apply on opensuse if they don't have that check in their version of overlayfs. On ubuntu FILESYSTEM_MAX_STACK_DEPTH is 2 so mounting a stack a b c d will fail when you go to mount d.

@kstaken Mmmm... I only tried with 3 layers so you may be right base + 1 + 1 is still in the max range.

I need to check whether our kernel patches also have that limit.

@kstaken you are completely right

May 28 10:21:47 piscola kernel: overlayfs: maximum fs stacking depth exceeded

I oversaw that. Documentation says "The lower filesystem can even be another overlayfs", but did not mention any limits. ARGH.

I would guess for docker the main usecase would be to have an image and keep the changes of the image separate, and then commit them to another image. Are more layers at the same time useful?
Because then we could introduce the limit to docker itself, when using overlayfs.

@kstaken I talked to Miklos (overlayfs author) as he work for us, and he confirmed that there is a limit to avoid a kernel stack overflow. He may look into adding arbitrary layers in the future, may be after it is included in the mainline kernel.
The question is whether docker really needs so many layers?

Contributor

kstaken commented May 28, 2013

I'm new to docker but I believe more layers at the same time is core to the way docker is intended to work.

The base image works but using a random image from the registry hits the limit.

docker run mbkan/lamp ps aux
2013/05/28 17:04:17 error: Error starting container b6a38d3e5dbc: Unable to mount using overlayfs

in syslog.

May 28 17:04:17 precise64 kernel: [111971.681350] overlayfs: maximum fs stacking depth exceeded

Is it pretty likely overlayfs is going to make it into the mainline kernel?

abohne commented Jun 6, 2013

+1 for EL support. I think targeting EL7 may be a more worthwhile effort than trying to move heaven and earth to get it working on EL6.

In case someone want to try, I have documented the steps for fedora 18 at http://neh.al/?p=1

NuxRo commented Jul 5, 2013

+1 for EL6 support. Also if you need alternative kernels you can contact elrepo.org who specialises in alt kernels and extra drivers for RHEL and clones.

+1 also for using a properly supported filesystem such as btrfs.

beders commented Jul 9, 2013

+1 for Redhat/CentOS support

sontek commented Jul 13, 2013

+1 For Redhat support

+1 for RH / Centos 6.4 support

jdoss commented Jul 18, 2013

+1 for RH / Centos 6 support

+1 for RHEL/CentOS 6+ support

+1 For RHEL/Centos 6.x support

Contributor

tamsky commented Jul 24, 2013

++ for Centos6/RHEL6 support

gbraad commented Jul 26, 2013

+1

+1

Absolutely +1. Docker isn't all that useful when the majority of commercial development happens on RHEL/Centos/Fedora/Oracle Linux based distros.

devinus commented Jul 30, 2013

👍

Contributor

thijsterlouw commented Jul 31, 2013

+1 otherwise our operations guys will block us from working with Docker. Need a clear upgrade path to RHEL7 and preferably on RHEL6.

+1 this is probably my highest item on wishlist.. several very large
organizations I consult with would adopt docker if we can get RHEL 6
support.
I'm keen to help out on this.

On Wed, Jul 31, 2013 at 9:53 PM, Thijs Terlouw notifications@github.comwrote:

+1 otherwise our operations guys will block us from working with Docker.
Need a clear upgrade path to RHEL7 and preferably on RHEL6.


Reply to this email directly or view it on GitHubhttps://github.com/dotcloud/docker/issues/172#issuecomment-21863959
.

hura commented Jul 31, 2013

+1

+1 For Redhat support

Contributor

mfojtik commented Aug 2, 2013

+1

+1 for RHEL 6 - that's my employer's approved linux server distro - ubuntu & others are non starters in our data center.

blalor commented Aug 3, 2013

+1 for RHEL/CentOS.

voltagex commented Aug 5, 2013

Can people please let me know if using a newer kernel from elRepo is not an option for them?

All the banks i consult at on rhel 5.6-8 just starting to move over to 6.x
I expect this is fairly common for large companies to lag similarly.

Sent from my iPhone

On 5 Aug, 2013, at 9:34, Adam Baxter notifications@github.com wrote:

Can people please let me know if using a newer kernel from elRepo is not an option for them?


Reply to this email directly or view it on GitHub.

mhitza commented Aug 5, 2013

+1 Fedora

I just did a write-up on how to install docker on CentOS / RHEL 6. The kernel Jérôme posted works great, you just need to remove the kernel-firmware package first.

http://blog.rage.net/2013/08/04/installing-docker-on-centos-6/

Jérôme, any chance of publishing a newer kernel RPM? All the docker docs say we should be using 3.8. I'll update my blog post with the newer kernel if one comes available.

voltagex commented Aug 5, 2013

I think the issue is big organisations won't want to shift kernel versions.

gbraad commented Aug 5, 2013

Ideally no alternative kernels should be used: people use Red Hat and buy a support contract for a reason. Installing an alternative, non-supported kernel, undermines the idea of running such a distribution. And please, do not compare Fedora and RHEL (CentOS, Scientific Linux, etc) to be the same. For Fedora the situation is different (more experimental in nature and not often deployed in a production environment)

Agreed. Will need to work with whatever we can in this case.

Sent from my iPhone

On 5 Aug, 2013, at 12:09, Gerard Braad — 吉拉德 notifications@github.com wrote:

Ideally no alternative kernels should be used: people use Red Hat and buy a support contract for a reason. Installing an alternative, non-supported kernel, undermines the idea of running a distribution.


Reply to this email directly or view it on GitHub.

NuxRo commented Aug 6, 2013

What's with the grsec dep?

Contributor

jpetazzo commented Aug 6, 2013

I build our kernels with GRSEC.

Yes :) I think LVM is just as viable as BTRFS as a storage backend. I would like to support them both with the plugin API.

@solomonstre
@getdocker

On Thu, Aug 8, 2013 at 12:55 PM, Brian Pitts notifications@github.com
wrote:

Has anyone evaluated using LVM's thinly-provisioned snapshot volumes for docker?
https://access.redhat.com/site/documentation/en-US/Red_Hat_Enterprise_Linux/6/html/Logical_Volume_Manager_Administration/thinly-provisioned_snapshot_volumes.html

https://fedoraproject.org/wiki/Features/ThinProvisioning

Reply to this email directly or view it on GitHub:
dotcloud#172 (comment)

sciurus commented Aug 13, 2013

I have a spec file for linux 3.10 with aufs at sciurus/docker-rhel-rpm. I haven't had much time to play with it yet, but this kernel and the lxc userspace (spec file also included) seem to get docker working on CentOS 6. I thought I'd also need aufs-utils, but I didn't package them yet and docker didn't complain that they were missing.

Please take this for a spin and let me know if you see any problems. Even better, fix them and send me a pull request!

If it would be useful, I can try to build and distribute binary RPMs via the OpenSUSE Build Service.

pdf commented Aug 13, 2013

Has anyone evaluated using LVM's thinly-provisioned snapshot volumes for docker?

LVM snapshots degrade pretty horribly with depth, because unlike COW-based snapshots, references to changed data must be updated all the way up the snapshot stack on each change.

sciurus commented Aug 13, 2013

@pdf I think that's true for traditional LVM snapshots, but not for thinly-provisioned ones.

pdf commented Aug 13, 2013

@sciurus looks like you're right (from your link):

The previous implementation of snapshots did this by chaining together lookup tables, and so performance was O(depth). This new implementation uses a single data structure to avoid this degradation with depth. Fragmentation may still be an issue, however, in some scenarios.

Sorry for the noise. I don't know how well thin LVM volumes are tested for ENOSPC though.

In any case performance concerns over CoW systems (which all pay some sort of performance penalty, however small, including aufs) are usually toned down by judicious use of bind-mounted volumes for performance-sensitive parts of the filesystem.

@solomonstre
@getdocker

On Mon, Aug 12, 2013 at 10:54 PM, Peter Fern notifications@github.com
wrote:

@sciurus looks like you're right (from your link):

The previous implementation of snapshots did this by chaining together lookup tables, and so performance was O(depth). This new implementation uses a single data structure to avoid this degradation with depth. Fragmentation may still be an issue, however, in some scenarios.

Sorry for the noise.

Reply to this email directly or view it on GitHub:
dotcloud#172 (comment)

sciurus commented Aug 16, 2013

I've added a package of docker itself at sciurus/docker-rhel-rpm. It's extremely simple as I use the precompiled docker binary and reuse the upstart job from Ubuntu. I've started playing with OBS here but am running into problems.

Contributor

boffbowsh commented Aug 19, 2013

As I'm hamstrung by EL5, I'm currently messing around with chroot and unionfs as a way to get docker-like functionality when you 100% trust the code you're trying to run and this is the only way of doing it. UnionFS is working perfectly for me. So why hasn't it been mentioned in the conversation so far?

Contributor

jpetazzo commented Aug 19, 2013

@boffbowsh , excellent news. Question (since I'm absolutely not familiar with UnionFS): does it support multiple layers? Or just one?

Contributor

boffbowsh commented Aug 19, 2013

Yep, multiple layers. Example from my ghetto heroku clone:

unionfs -ocow $OVERLAY_DIR=RW:$SLUG_DIR=RO:$BASE_DIR=RO $MOUNT_DIR

sciurus commented Aug 19, 2013

@boffbowsh does the stock EL5 kernel include the UnionFS patches? I can't find the unionfs module or userspace tools on a sample RHEL5 system.

Contributor

jpetazzo commented Aug 20, 2013

It's based on FUSE, right?

On Mon, Aug 19, 2013 at 3:15 PM, Brian Pitts notifications@github.comwrote:

@boffbowsh https://github.com/boffbowsh does the stock EL5 kernel
include the UnionFS patches? I can't find the unionfs module or userspace
tools on a sample RHEL5 system.


Reply to this email directly or view it on GitHubhttps://github.com/dotcloud/docker/issues/172#issuecomment-22908813
.

sciurus commented Aug 20, 2013

@jpetazzo No, it doesn't use FUSE. Aufs actually began as a fork of it; see https://lwn.net/Articles/327738/

Contributor

boffbowsh commented Aug 20, 2013

The instructions I followed to get it up and running use FUSE. I figure UnionFS is "safe" as it seems like the majority of LiveCDs use it.

nareshv commented Aug 21, 2013

I have created a centos-6.x repository for docker.io with aufs. Thanks to @sciurus for the kernel spec file.

NuxRo commented Aug 21, 2013

Nareshv, that's great, thanks for your effort!

Thanks for your kernel RPM. Awesome work. I've been testing it on RHEL 6.4 and it works like a charm.

Also, for those interested, I published a Chef cookbook that takes care of installing and configuring Docker.

https://github.com/failshell/chef-docker

chenryn commented Aug 24, 2013

@nehaldattani I try to do the same thing on Fedora19 but when I compile kernel-3.10 with aufs3 patch(3.10 branch too), it failed. Are you try this on fc19?

@chenryn : At what point it failed ? Can you load aufs module ?

chenryn commented Aug 26, 2013

@nehaldattani
CHECK include/mtd (5 files)
/root/rpmbuild/BUILD/kernel-3.10.0+/user/include/linux/aufs_type.h:23: included file 'linux/limits.h' is not exported
/root/rpmbuild/BUILD/kernel-3.10.0+/user/include/mtd/inftl-user.h:9: included file 'linux/types.h' is not exported
/root/rpmbuild/BUILD/kernel-3.10.0+/user/include/asm-generic/fcntl.h:4: included file 'linux/types.h' is not exported
/root/rpmbuild/BUILD/kernel-3.10.0+/user/include/mtd/mtd-abi.h:23: included file 'linux/types.h' is not exported
make[5]: *** [/root/rpmbuild/BUILD/kernel-3.10.0+/usr/include/linux/.check] Eror 123
make[4]: *** [linux] Error 2

@chenryn : Are you using code from kernel.org or the code has been extracted from SRPM ?

chenryn commented Aug 26, 2013

@nehaldattani I had both tried the kernel code from git branch 3.10, and from kernel-3.10.7-200.fc19.src.rpm. Got the same result

@chenryn : I have created a new F19 vm and applied the aufs patches on kernel 3.10 ( code from kernel.org) and just completed building of kernel rpm with aufs module.

In your case, it seems that all aufs header files are not included properly.

I also found a typo in my post. ( Though you will get a message "cp:ommiting directory .....)

Read "cp include/uapi/linux/ ../linux-2.6/include/uapi/linux/" as "cp include/uapi/linux/aufs_type.h ../linux-2.6/include/uapi/linux/"

Please copy header files as mentioned in my post ( exception is above mentioned file) and give it a try.

Good Luck !!!

sciurus commented Aug 26, 2013

@chenryn have you looked at sciurus/docker-rhel-rpm? If your goal is to build a 3.10 kernel with aufs support, you can use my steps there. I've added some more information about this to the README.

jlk commented Sep 3, 2013

+1 - Would love to see solid EL6 support. Curious to see where this goes...

mztriz commented Sep 17, 2013

+1 I would like to see some RHEL support as well.

+1, yes please.

vannenc commented Sep 19, 2013

+1, for CentOS/RedHat

Contributor

bflad commented Sep 19, 2013

That is awesome news.

Is there any information about where this plugin API is being designed? I don't see any relevant branches in the Github repository. I'm very interested in the possibility of a ZoL plugin as an alternative to btrfs and thinly-provisioned device-mapper, but if a btrfs plugin is already in development, it would be nice to be able to see that code to see what it would look like to make a ZoL version.

👍

+1 for Red Hat support

+1 centos 6, and 5 if you can get it

Contributor

goldmann commented Nov 8, 2013

A blog post describing the details about RHEL/Fedora support for Docker: http://community.redhat.com/adventures-in-dockerland/

sciurus commented Nov 26, 2013

Now that

  • Docker 0.7 is released and supports devicemapper thin provisioning as an alternative to aufs
  • Docker has been packaged in Fedora and EPEL

I think this issue can be closed.

Contributor

goldmann commented Nov 26, 2013

It's packaged, but not pushed to the repos yet. Please hold on, we'll be there soon!

blalor commented Nov 26, 2013

Will Docker 0.7 work on CentOS 6? I think that's still on kernel 2.6, but the Docker docks say 3.8 (?) is required.

NuxRo commented Nov 26, 2013

It works nicely:
yum --enablerepo=epel-testing install docker-io libcgroup
service cgconfig start
service docker start

blalor commented Nov 26, 2013

Excellent. yum resolutely refused to find docker-io in epel-testing, but I used “yum localinstall” with the URL to the package IN THE EPEL-TESTING REPO and it works a treat. Yum even found lxc and lxc-libs in epel-testing, just not docker-io. I just started the centos container on CentOS.

MY GOD, IT’S FULL OF STARS!

blalor commented Nov 27, 2013

Er, not quite. The docker-io service in epel-testing starts with -b none, so networking isn't available. Attempting to start docker without that flag causes it to fail.

[root@vagrant-centos63 tmp]# /usr/bin/docker -D -r=false -d
2013/11/26 19:25:07 WARNING: You are running linux kernel version 2.6.32-279.19.1.el6.x86_64, which might be unstable running docker. Please upgrade your kernel to 3.8.0.
[/var/lib/docker|f4b03b56] +job initapi()
[/var/lib/docker|f4b03b56.initapi()] Creating server
[debug] driver.go:84 Error loading driver aufs: AUFS was not found in /proc/filesystems
[debug] deviceset.go:475 Generated prefix: docker-253:0-404404
[debug] deviceset.go:478 Checking for existence of the pool 'docker-253:0-404404-pool'
[debug] deviceset.go:237 loadMetadata()
[debug] deviceset.go:242 loadMetadata END
[debug] runtime.go:642 Using graph driver devicemapper
[debug] network.go:150 Creating bridge docker0 with network 172.17.42.1/16
[/var/lib/docker|f4b03b56] -job initapi() = ERR (Error creating bridge: operation not supported)
2013/11/26 19:25:07 initapi: Error creating bridge: operation not supported

NuxRo commented Nov 27, 2013

Right, so you either need to create a bridge (docker0) by yourself and specify it in /etc/sysconfig/docker or go for a newer kernel.

Kernel-lt from Elrepo works fine at a first glance.

NuxRo commented Nov 27, 2013

Until docker-io makes a comeback in epel-testing it can be downloaded from http://koji.fedoraproject.org/koji/packageinfo?packageID=17053

blalor commented Nov 27, 2013

I'm working on a sample Vagrant box and bootstrapping script for CentOS. https://github.com/blalor/vagrant-centos64-docker It's not working yet, but there has been some good discussion on docker-dev today. You don't really need an RPM as you can just grab the docker binary. My Vagrant config still doesn't work (lxc incompatibility?) but I hope to get back to it later tonight to get it working.

Contributor

bflad commented Dec 3, 2013

Just to follow up with this thread, the bridging and networking on CentOS 6 should be fixed along with the libcgroup/cgconfig fixes in docker-io packages as of last few days. According to Red Hat Bugzilla 1000662 and Fedora build system, the latest docker-io package (docker-io-0.7.0-14) is being pushed to EPEL 6 stable right now. Fedora build system is showing FC19/FC20 being pushed to updates stable repositories shortly as well.

Great work everyone!

blalor commented Dec 3, 2013

Great news!! Last I looked the docker-io package didn't set up the docker0 bridge. Does it now do that, or is that an additional step required to get Docker containers to have network access?

Contributor

bflad commented Dec 3, 2013

All handled without effort with latest package, at least from what I can tell. :)

[root@docker-centos-6 ~]# rpm -ihv http://kojipkgs.fedoraproject.org//packages/docker-io/0.7.0/14.el6/x86_64/docker-io-0.7.0-14.el6.x86_64.rpm
Retrieving http://kojipkgs.fedoraproject.org//packages/docker-io/0.7.0/14.el6/x86_64/docker-io-0.7.0-14.el6.x86_64.rpm
Preparing...                ########################################### [100%]
   1:docker-io              ########################################### [100%]
[root@docker-centos-6 ~]# cat /etc/sysconfig/docker
other_args=""
[root@docker-centos-6 ~]# service docker start
Starting cgconfig service:                                 [  OK  ]
Starting docker:                                     [  OK  ]
[root@docker-centos-6 ~]# ps aux | grep docker
root      5111  1.0  1.2 542220 12416 pts/0    Sl   03:21   0:00 /usr/bin/docker -d
[root@docker-centos-6 ~]# docker run -d -p 9999:9999 busybox /bin/nc -l -l -p 9999
Unable to find image 'busybox' (tag: latest) locally
Pulling repository busybox
e9aa60c60128: Download complete
661e7e01438e581b0acce27ea15856b824461cc64914e46b8d374e954013188b
[root@docker-centos-6 ~]# netstat -ntlp | grep 9999
tcp        0      0 :::9999                     :::*                        LISTEN      5111/docker
[root@docker-centos-6 ~]# telnet ::1 9999
Trying ::1...
Connected to ::1.
[root@docker-centos-6 ~]# brctl show
bridge name bridge id   STP enabled interfaces
docker0   8000.000000000000 no
Contributor

goldmann commented Dec 3, 2013

@blalor It's fixed in the latest update.

All--

From now on please report all issues regarding docker-io EPEL package in the Red Hat Bugzilla. Thanks!

I'll update the docs with instruction on how to install Docker on CentOS/RHEL.

I guess we can close this issue now.

Contributor

jamtur01 commented Dec 26, 2013

Let's close it! :) Thanks @goldmann!

@jamtur01 jamtur01 closed this Dec 26, 2013

mrunalp pushed a commit to mrunalp/docker that referenced this issue Jul 20, 2016

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment