On 29 Sep 2005, at 12:50, Steve Traylen wrote:
> On Thu, Sep 29, 2005 at 11:01:14AM +0100 or thereabouts, Tim Barrass
> wrote:
>> Sorry John, that reminds me-- I wanted to make a comment but the
>> discussion (and the mic) went elsewhere, and then I ran out of time
>> and
>> had to go.
>>
>> You asked if it might be possible for the experiments to pack their
>> agents/services for easy install via rpm. I would say this is
>> possible,
>> but with issues:
>>
>> Taking the example of PhEDEx, our local CMS guys take time fiddling
>> small glue scripts that allow PhEDEx to access whatever local
>> combination of storage tech and transfer tool exist. We find that
>> there
>> is significant variation in site setups that mean that we can't simply
>> provide 2 or 3 out-of-the-box solutions.
>
> Move forward a bit and assuming all sites have DCache or DPM which do
> of
> course vary a little still what else is a local configuration?
This is kind of time-frame coupled. In the short term we still need to
do things like issue 'ls' or 'cksum' to check that a file exists after
transfer, and has the right size. We also need to burrow into various
tape systems to issue baroque commands to work out whether files have
been migrated yet.
As time goes on I hope that our feedback with tools providers makes the
tools more reliable so we don't need to do this. FTS for example will
take a lot of the worry for this off our hands... when it works and
we've worked out how to lay PhEDEx over the top in detail. Although we
won't be able to use FTS for US transfers, so we'll still need to
support some equivalent structure. One critical issue though is knowing
for "sure" that a file has been migrated-- in principle a tape SRM
might be reliable enough that if it gets as far as the stage disk it's
as good as migrated... we are also beginning to push a lot of the
responsibility for managing efficient stages etc onto source SRMs as
they mature.
After that it's down to more easily set local configuration-- where to
put files, hostnames etc.
And sorry; at this point I can't quantify short- or long-term. At some
stage though I'd like to kick off a rationalisation of the lower levels
of this system. At the moment everyone involved is up to their eyeballs
in it though.
I get the feeling that other CMS services: PubDB for example, or our
publication and validation scripts, are more passive, and need to do
less burrowing into the system.
Sorry for the loose answer :/
Tim
> Steve
>>
>> I just raise this is it might be a complicating issue for easy
>> installation of services.
>>
>> Thanks,
>> Tim
>>
>> On 29 Sep 2005, at 10:15, Gordon, JC (John) wrote:
>>
>>> Considering the heat in the VO Box discussion on this list I was
>>> surprised that more wasn't raised during the discussion on Tuesday.
>>>
>>> This is probably your last chance to get the result we want.
>>>
>>> After this workshop, complaints within GRidPP will not carry much
>>> weight.
>>>
>>> John
>
> --
> Steve Traylen
> [log in to unmask]
> http://www.gridpp.ac.uk/
|