Reverting to what? -1 or -/+0? CentOS6 and openSUSE are still *very* important platforms we should support. And as we agreed, I will run build and tests on all openSUSE and Fedora once a week. I can also handle CentOS6.
If we add CentOS5 that's great, but I would rather close that vote first and then add centos5 when we get the resource up and running. I don't want to take the chance to loose support for CentOS6 or openSUSE. It would be a lot more work to re-introduce them than keeping them working. On 09/09/2011 12:26 PM, Andrew Bayer wrote: > I'm reverting my +1 - OSUOSL doesn't yet have images for CentOS 6, OpenSUSE, > Fedora. I personally believe we should only officially support those > platforms which we can actually do full end-to-end testing on, and our > end-to-end testing is most definitely moving to OSUOSL, so... > > My proposal would be CentOS 5 and Lucid for 0.2.0. Roman's working on more > images for use at OSUOSL, so hopefully we can expand the list for 0.3.0, but > for now, I think we're gated by what platforms we can actually test on. > > A. > > On Fri, Sep 9, 2011 at 11:43 AM, Bruno Mahé <[email protected]> wrote: > >> On 09/09/2011 03:48 AM, Steve Loughran wrote: >>> On 06/09/11 21:44, Bruno Mahé wrote: >>>> Hi, >>>> >>>> Given no one has objected to the previous thread, I am calling a vote. >>> I'm not sure if I've missed the vote here so >>> >> This is the last day to vote I believe. >> And If I understand correctly, you are -1 overall. >>>> So let's vote on adopting the following set of platforms: >>>> * Latest released CentOS (currently 6.0) >>> =0. I've heard of problems related to glibc and other RHEL quirks >>> https://issues.apache.org/jira/browse/HADOOP-7156 >>> >>> The alternative would be the 5.7+ branches, which are what people tend >>> to run in production >>> >> Right now we are still setting up the project and working on getting >> resources. >> In the mean time, I am for example running some build and tests on my >> own VMs. >> >> So for now I would rather have us focus on a wider range of >> distributions so we can ensure BigTop exercise most packaging paths. And >> then, when we can get access to more resources we would add support for >> CentOS5. >> Having more frequent releases of Hadoop would also help (CentOS5 is >> getting old and we can't patch anything for build or security issues). >> >> I am also planning on setting up builds against trunk of each projects >> so we can potentially help stabilizing them and ensuring they work >> properly when they get released. This would make it easier to support >> old GNU/Linux distributions. >> >>>> * Latest released Ubuntu LTS (currently 10.04) >>> +1, even if the first thing you have to do on a 10.04 installation is >>> delete the hostname -> loopback mapping in /etc/hosts >>> >>>> * Latest released OpenSUSE (currently 11.4) >>> +1 if someone wants to do it >>> >>>> * Latest released Fedora (currently 15) >>> I'm a -1 on this because it's too bleeding edge, it's not LTS, and you >>> will end up debugging fedora problems. >>> >> I usually rebuild CDH on fedora + openJDK, and now BigTop on openJDK for >> my personal usage and so far have hit only one issue which is already >> fixed in 0.22, 0.23 and trunk. >> So I am not really concerned about its bleeding edge. Actually I am more >> concerned about old distributions (CentOS5, Debian Lenny...) than the >> bleeding edge ones. >> >> I also noticed a few tickets on JIRA about getting hadoop working on >> Fedora. >> >> >>>> This means we can't check in any patch that will break any of these >>>> OSes. >>>> It also means we can't upgrade a component of BigTop if that component >>>> cannot be built or tested on all of these OSes. >>>> >>>> >>>> This does not preclude anyone to add support for or maintain any other >>>> platform, but this would not be a requirement for BigTop. >>>> >>>> >>>> The vote is open for 72 hours, and is open to anyone on the committers >>>> or mentors list at http://incubator.apache.org/projects/bigtop.html. >>>> >>>> >>>> Thanks! >>>> >>> >>> >>
