The current implementation of HTTP(s) health check is based on curl.
According to the document of curl

>Since 7.47.0, the curl tool enables HTTP/2 by default for HTTPS
connections.

So I think it should be supported if the curl version in your Mesos Agent
is higher that 7.47. But I have not yet try this.

On Sat, Sep 3, 2016 at 12:23 AM, Aaron Wood <aaronjw...@gmail.com> wrote:

> Since you mentioned that you're working on supporting HTTPS health checks
> I'm curious if there are any plans to support HTTP/2 over TLS (or even over
> plain HTTP). I would think that using HTTP/2 for any communication that
> happens in Mesos would provide a nice improvement in heavy load situations.
>
> On Fri, Sep 2, 2016 at 10:59 AM, haosdent <haosd...@gmail.com> wrote:
>
> > Hi, dear friends. @alexr and I are working on supporting HTTP(s)/TCP
> Health
> > Check in Mesos.
> > We have finished and committed some initial works. But if you use the old
> > protobuf definition of
> > `HealthCheck` to implement HTTP health check in your custom executor
> > before, our changes recently would
> > break it.
> >
> > The change of the protobuf definition of `HealthCheck` is
> >
> > ```
> >  message HealthCheck {
> >  +  enum Type {
> >  +    UNKNOWN = 0;
> >  +    COMMAND = 1;
> >  +    HTTP = 2;
> >  +    TCP = 3;
> >  +  }
> >  +
> >  -  message HTTP {
> >  +  message HTTPCheckInfo {
> >  +    optional string scheme = 1;
> >  -    required uint32 port = 1;
> >  +    required uint32 port = 2;
> >  -    optional string path = 2 [default = "/"];
> >  +    optional string path = 3;
> >  -    repeated uint32 statuses = 4;
> >     }
> > ...
> >  +  optional Type type = 8;
> >  -  // HTTP health check - not yet recommended for use, see MESOS-2533.
> >  -  optional HTTP http = 1;
> >  +  optional HTTPCheckInfo http = 1;
> > ...
> >   }
> > ```
> >
> > Noted that we add a field `type` to specific the health check type and
> use
> > `HTTPCheckInfo` instead of `HTTP`.
> > As I know, Mesos didn't support HTTP health check before 1.0 and it is
> > supposed to not used.
> >
> > But thanks to @swsnider to report the issues recently, user may implement
> > the custom executor with
> > HTTP health check. So I am writing this email to check if anyone
> > implemented HTTP health check in custom executor
> > like @swsnider and if you depend on the old protobuf definition of
> > `HealthCheck` heavily.
> > If so, how many month your need for the deprecation cycle of this?
> >
> > Any concerns and questions are appreciated, thanks a lot!
> >
> > --
> > Best Regards,
> > Haosdent Huang
> >
>



-- 
Best Regards,
Haosdent Huang

Reply via email to