[ClusterLabs Developers] bundle/rkt implementation in pacemaker
Oyvind Albrigtsen
oalbrigt at redhat.com
Tue Jul 18 03:02:34 EDT 2017
On 18/07/17 08:18 +0200, Valentin Vidic wrote:
>On Mon, Jul 17, 2017 at 03:43:22PM -0500, Ken Gaillot wrote:
>> See the xml/Readme.md file -- I won't say all our documentation is
>> helpful, but that one is pretty good. ;-) Basically, you create a
>> resources-2.10.rng and leave everything else alone, and the makefile
>> will take it from there.
>>
>> However, there's a catch. Pacemaker's internal XML revision number
>> implementation is a floating-point number, so 2.10 == 2.1. It's been on
>> the to-do list to fix that for a while, and it's gained new urgency
>> since we hit 2.9, but it hasn't happened yet. Definitely a high priority
>> :-) For now, for testing purposes, just modify the 2.9 file.
>>
>> Regarding the patch, I'm thinking we should add a flags bitmask to
>> container_variant_data_t, and container_unpack() could set a flag for
>> the container type. That way, you don't have to do first_named_child()
>> every time you need to check the type, and you can check the type in
>> allocate_ip() so it doesn't break the docker code.
>
>Thanks for the review, I'll try that out. Can you give me a hint about:
>
> create_nvp(xml_obj, "mount_points", dbuffer);
>
>I don't see mount_points being used anywhere, do I need it in the rkt
>part of the code?
mount_points might be good to have for rkt as well, as it creates
directories on the host server that the container will mount:
https://github.com/ClusterLabs/resource-agents/blob/master/heartbeat/docker#L109
>
>--
>Valentin
>
>_______________________________________________
>Developers mailing list
>Developers at clusterlabs.org
>http://lists.clusterlabs.org/mailman/listinfo/developers
More information about the Developers
mailing list