Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Sign In with Google Sign In with OpenID Sign In with Twitter
Looking for community input: EL6 vs. EL7
  • pmyerspmyers +1 -1 (+1 / -0 )
    Posts: 65

    As many folks in the RDO community are aware, RDO was created to provide a stable distribution of OpenStack on top of Enterprise Linux operating system distributions. It also provides a way to get the latest releases of OpenStack on top of stable versions of Fedora.

    For the last few OpenStack release cycles, we have focused on providing OpenStack on top of RHEL 6 and its derivatives. In order to do this, we've had to provide some updates in RDO for things like the kernel and other userspace packages (iproute, openvswitch and some python libraries) that were too new or disruptive to include in base RHEL 6. In short, trying to get cutting edge software (OpenStack) running on a base platform that is already several years old, has at times been challenging.

    In the future, there will of course be a new version of RHEL (assuming sequential numbering, this would be RHEL 7). RHEL 7 would be based on a recent version of Fedora, and so would be starting with a much more modern kernel and package set. This would give RDO a significant boost in terms of keeping up with the latest and greatest features from upstream OpenStack.

    However, we are aware that there are users who may want to stay on the EL6 variants after RHEL 7 and EL7 variants are released. So, we would like to hear from you, the community, what your preferences would be. There may not be a simple EL6 or EL7 decision, but any input you have will help us to determine how quickly we should move from EL6 to EL7 and whether or not it is imperative to support an OpenStack upstream release on both platforms simultaneously.

    There are three likely models for how we deal with the shift from EL6 to EL7 in RDO. Please consider each of these and let us know what you think of the pros/cons for each:

    Immediate Cut Over:

    • Support RDO 'X' on EL6
    • EL7.0 is released
    • Support RDO 'X+1' on EL7.0

    Delayed Cut Over:

    • Support RDO 'X' on EL6
    • EL7.0 is released
    • Support RDO 'X+1' on EL6
    • EL 7.1 is released
    • Support RDO 'X+2' on EL7.1

    Note: You could replace 7.1 above with any arbitrary minor release of EL7, depending on how long of a delay is desired.

    Simultaneous Support:

    • Support RDO 'X' on EL6
    • EL7.0 is released
    • Support RDO 'X+1' on EL6 and EL7.0 (two separate packaging streams needed)
    • EL7.1 is released
    • Support RDO 'X+2' on EL7.1

    Note: You could have multiple releases of RDO where both EL6 and EL7 are both supported.

    Again, we really do want to hear from you, the community, on your preferences. Looking forward to hearing and discussing your input!

    Cheers,

    Perry (on behalf of the RDO team)

  • red_trelared_trela +1 -1 (+1 / -0 )
    Posts: 47

    I tend to think RDO relates to RHOS as Fedora does to RHEL. That said, I'd always expect the latest and greatest and only minimal to no support for older technology. True bleeding edge(TM).

    At the same time, the RHEL ecosystem is a bit more complicated in some respects than Fedora's. Important to this discussion: the release of RHEL 7 does not necessarily mean 7 is released at the virtually same time or even within short time.

    So I'd want to keep as close to the "immediate cut over" model as possible but take the above into account and go with a slightly modified "simultaneous support" model, Simultaneous Support until Base OS Parity:

    • Support RDO 'X' on EL6
    • RHEL 7 is released
    • Support RDO 'X+1' on EL6 and EL7 (two separate packaging streams needeD)
    • CentOS 7, SL 7 and maybe other considered-important derivatives based on RHEL 7 are released
    • (Optionally: drop RDO 'X+1' support on EL6 - in this scenario, 'X+1' would only be used on EL6 by people waiting to switch to EL7 ASAP)
    • Support RDO 'X+2' on EL7

    Note, that by "EL" I always mean RHEL-and-derivatives unless I write RHEL, etc. And minor releases don't matter above as once a derivative is has released a new major version, they're very good at keeping up.

  • marafamarafa +1 -1
    Posts: 85

    @pmyers

    was going to write up a long post but lets keep this short: Simultaneous Support and thanks for asking

    thank you

  • pmyerspmyers +1 -1
    Posts: 65

    Thanks @red_trela and @marafa for your input! One question... It's going to be a much larger packaging effort to handle both EL6 and EL7 simultaneously. It would be great if we could start to get community participation in the packaging efforts. Is that something that both of you would entertain? :)

  • dnearydneary +1 -1
    Posts: 262

    Hi,

    Would there be any interest in something like:

    • RDO X maintained on RHEL 6
    • RHEL 7 released
    • RDO X+1 maintained on RHEL 7, but we provide access for interested parties to do RHEL 6 packaging

    It's a half-way house between instant switch-over and dual streams - Red Hat would only have the resources to build one stream, but we could provide a distribution channel for RPMs of the second stream - there would be some details to work out there related to build infrastructure, testing, package signing, etc - but would something like that work? If so, who would be interested in maintaining the packages?

    Cheers,
    Dave.

  • From where I sit and the corporates I deal with the only option that will really fly is a modified simultaneous support.....

    Support RDO 'X' on EL6
    EL7.0 is released with RDO 'X'
    Support RDO 'X+1' on EL6 and EL7.0 (two separate packaging streams needed)
    Support RDO 'X+2' on EL7.x 
    

    I think one of the critical factors is that RDO Tracking upstream OpenStack means it will have a much shorter release cycle than the underlying OS. I must admit I haven't fully thought through the issue.

  • Keith_ChambersKeith_Chambers +1 -1 (+1 / -0 )
    Posts: 1

    Definitely should not drop support for EL6 until there is a community rebuild ... i.e., CentOS 7.

  • marafamarafa +1 -1
    Posts: 85

    @pmyers - i must apologise but i must decline the honor. what "free" time i have is limited

  • I'd be interested in helping to maintain an EL6 branch @pmyers and @dneary

  • kfox1111kfox1111 +1 -1
    Posts: 12

    Yeah, RHEL 6.4 is kind of awkward because to really use it, you have to drop support by using unsupported kernel/bits.

    Rumor is 6.5 will support everything needed to make OpenStack work out of the box so the pressure to get to RHEL 7 will be off if that is the case. I'm guessing you can't talk about that though. :)

    The fast majority of nodes in a cloud are just compute nodes. They have minimal requirements and will probably get a good performance boost right away by going to newer hypervisors. So maybe continue to target RHEL 6 for another release after RHEL 7 but parallel support compute nodes only on RHEL7. This allows quickly upgrading most of the cluster to 7, and then a more careful upgrading of the rest of the bits.

    If the rumor is not true, then getting to RHEL7 as soon as possible so that it can be supported again would be preferable.

  • @pmyers well, I could help somewhat. For a limited time. Unfortunately, there's really a lot of constraints here right now. And I figure we're speaking about Icehouse or J-release, right? Because I have no idea yet how my life will look like in 2014 (since I'm going to move to Japan).

  • carltmcarltm +1 -1
    Posts: 17

    I'd like to see simultaneous support until RDO has been verified on CentOS 7.