corosync?

corosync?

[Brad Bishop]

Is anyone using corosync/pacemaker on a BMC?  Is that a crazy idea? :-)

thx - brad

Re: corosync?

[Patrick Williams]

[-- Attachment #1: Type: text/plain, Size: 240 bytes --]

On Tue, Mar 24, 2020 at 11:33:23AM -0400, Brad Bishop wrote:
> Is anyone using corosync/pacemaker on a BMC?  Is that a crazy idea? :-)
> 
> thx - brad

Can you elaborate on what you'd be trying to accomplish?

-- 
Patrick Williams

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: corosync?

[Brad Bishop]

at 12:06 PM, Patrick Williams <patrick@stwcx.xyz> wrote:

> On Tue, Mar 24, 2020 at 11:33:23AM -0400, Brad Bishop wrote:
>> Is anyone using corosync/pacemaker on a BMC?  Is that a crazy idea? :-)
>>
>> thx - brad
>
> Can you elaborate on what you'd be trying to accomplish?

Thanks Patrick.  Yes I will definitely do that at some point.  I’m not  
ready to do this right now though simply because I myself am not sure what  
I am trying to accomplish.

In this moment I’m just curious if anyone has any experience with it, or  
similar tools or frameworks, in a "BMC-like" distributed system, and the  
sorts of problems that were solved with it/them.

thx - brad

Re: corosync?

[Richard Hanley]

[-- Attachment #1: Type: text/plain, Size: 1663 bytes --]

We looked into getting Envoy as a proxy for a distributed BMC network.  The
idea was that Envoy could be used for discoverability and creating long
lived authenticated channels with HTTP/2.  That work got mostly shelved
because it was really hard to get the build system to work on 32-bit arm
(not to mention getting it to work with bitbake).

I'm not too familiar with corosync.  I remember reading a bit about it when
Vishwa mentioned aggregating BMCs a few months ago.  It looks like it
should be relatively easy to build (at least compared to envoy).

Unfortunately we never got a chance to see how much cpu usage is used when
Envoy is run on a Poleg.  In terms of sheer craziness, corosync isn't any
less crazy than what we were thinking about, and I was relatively
optimistic that it could work on current generation hardware.

- Richard

On Tue, Mar 24, 2020 at 9:22 AM Brad Bishop <bradleyb@fuzziesquirrel.com>
wrote:

> at 12:06 PM, Patrick Williams <patrick@stwcx.xyz> wrote:
>
> > On Tue, Mar 24, 2020 at 11:33:23AM -0400, Brad Bishop wrote:
> >> Is anyone using corosync/pacemaker on a BMC?  Is that a crazy idea? :-)
> >>
> >> thx - brad
> >
> > Can you elaborate on what you'd be trying to accomplish?
>
> Thanks Patrick.  Yes I will definitely do that at some point.  I’m not
> ready to do this right now though simply because I myself am not sure
> what
> I am trying to accomplish.
>
> In this moment I’m just curious if anyone has any experience with it, or
> similar tools or frameworks, in a "BMC-like" distributed system, and the
> sorts of problems that were solved with it/them.
>
> thx - brad
>

[-- Attachment #2: Type: text/html, Size: 2182 bytes --]

Re: corosync?

[Brad Bishop]

at 8:37 PM, Richard Hanley <rhanley@google.com> wrote:

> We looked into getting Envoy as a proxy for a distributed BMC network.   
> The idea was that Envoy could be used for discoverability and creating  
> long lived authenticated channels with HTTP/2.  That work got mostly  
> shelved because it was really hard to get the build system to work on  
> 32-bit arm (not to mention getting it to work with bitbake).
>
> I'm not too familiar with corosync.  I remember reading a bit about it  
> when Vishwa mentioned aggregating BMCs a few months ago.  It looks like  
> it should be relatively easy to build (at least compared to envoy).
>
> Unfortunately we never got a chance to see how much cpu usage is used  
> when Envoy is run on a Poleg.  In terms of sheer craziness, corosync  
> isn't any less crazy than what we were thinking about, and I was  
> relatively optimistic that it could work on current generation hardware.
>
> - Richard

Thanks for the reply Richard!

-brad