Re: Proposal for Remote Execution

[Jim MacArthur <jim macarthur codethink co uk>]

On 18/04/18 17:01, Tristan Maat wrote:
> Forwarding this little thread for completeness, I
> look to have forgotten to CC the ML:
>
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>
> Hi Tristan,
>
> On Wed, 2018-04-18 at 16:19 +0100, Tristan Maat wrote:
> > > CAS Artifact Cache
> > > ~~~~~~~~~~~~~~~~~~
> > > As a first phase I'm proposing to introduce an artifact cache based 
> > on CAS
> > > as defined in the Remote Execution API. This can completely replace 
> > the
> > > existing tar and OSTree artifact cache backends with a single 
> > cross-platform
> > > implementation.
> >
> > Considering this should just plug snuggly into the existing cache 
> > interface, I
> > don't expect this to be a problem, but to be on the safe side - I'm 
> > assuming
> > the WIP cache expiry methods can be implemented in a CAS-based cache 
> > as well?
>
> Yes, the structure of the CAS artifact cache is very similar to the
> OSTree artifact cache except that there are no commit objects, refs
> directly point to Merkle tree hashes. I haven't implemented a purge
> method yet but I'm not expecting any significant differences with
> regards to cache expiration.
If we're planning to use the Bazel method then there will be linked 
objects in the CAS, e.g. build logs and build commands linked to a 
particular build action, more things than just the build result which we 
currently cache. We'll probably want to delete all the linked objects at 
the same time, which might require some more thought.