[Piuparts-devel] protocol changes for idle and arch

Holger Levsen holger at layer-acht.org
Mon Jun 4 13:16:58 UTC 2012


Hi Andreas,

On Montag, 4. Juni 2012, Andreas Beckmann wrote:
> master goes into idle state if it can't fulfill a reserve request
> idle state expires after some time (1 hour) to see if Packages has been
> updated

yes, but why is this extension to the protocol needed? If the master has 
nothing to reserve, it says so, and the slaves retries after some time. 

I like the simplistic protocol and to keep it this way, if possible.
 
> submitting a log (or any other modification of the logdb, e.g.
> rescheduling) clears the idle state and requires status recomputation on
> the next connection

/me nods (+so what, see above ;)
 
> the important part is that just querying the idle state is cheap (no
> loading of Packages and logDB)

so you can fail on "reserve" equally fast?!

> Just imagine you have 7 slaves running and about 50 sections in your
> config ... like me :-)

:) Are those 7 slaves the same arch?


cheers,
	Holger



More information about the Piuparts-devel mailing list