-
Notifications
You must be signed in to change notification settings - Fork 2
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add tables for map version 160812 for 3 and 3.8 T #1
Conversation
A new Pull Request was created by @namapane (Nicola Amapane) for branch master. @cmsbuild, @smuzaffar, @mrodozov, @iahmad-khan, @davidlange6 can you please review it and eventually sign? Thanks. external issue cms-sw/cmsdist#3094 |
I belatedly notice - 22k files (11k per map?).. can we not condense these somewhat? |
ok, i guess this is continuing an old design... find /cvmfs/cms.cern.ch/slc6_amd64_gcc530/cms/data-MagneticField-Interpolation/V01-00-00/MagneticField/Interpolation/data/grid_130503_3_8t_v9_small | wc |
right, currently in tag V01-00-00 we have 48K files with total size of 380MB |
These are for different map, ie for different nominal current, and
separately for RunI/RunII. There are also different version with more
and more refined models, and eventually these new ones we are preparing
now will replace some of the old ones.
If we badly need to reduce the overall number of files, we could remove
two sets that are not strictly needed (grid_1103l_071212_4t and
grid_120812_3_8t_v7_small).
Regarding condensing: a possibility that has been considered in the past
is to move the whole amount of data to the conditions db; this would
require some work and some performance studies for which manpower is
unfortunatley quite short.
Cheers,
Nicola
…On 13-Jun-17 10:11, Malik Shahzad Muzaffar wrote:
right, current in tag V01-00-00 we have 48K files with total size of 380MB
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#1 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AE7eUJMSBFKSBT7wu87djNHM1i6gDs2Vks5sDkQcgaJpZM4Ny4ks>.
|
I picked a recent nightly build and I find this under $CMSSW_DATA_PATH/data-MagneticField-Interpolation/V00-00-01/... and it works. |
On 6/23/17 9:32 AM, Nicola Amapane wrote:
I picked a recent nightly build and I find this under
$CMSSW_DATA_PATH/data-MagneticField-Interpolation/V00-00-01/... and it
works.
However, I also see that
$CMSSW_DATA_PATH/data-MagneticField-Interpolation/V00-00-00/... exists,
which is also populated. What is the reason for keeping both? In this
way a good fraction of this large number of files is distributed twice,
which does not seem to make a lot of sense.
the same area is used for distributions of all supported past releases.
…
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#1 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AEdcbpX-3dIjZVn-ZdVm_IyO-5NJW4dZks5sG-i3gaJpZM4Ny4ks>.
|
Yes but in this case V00-00-01 contains exactly the same files of V00-00-00, plus two sets in addition which do not interfere with supported past releases. There is really no point in duplicating these 48k files, couldn't we have added the new files and created a new set only when we really break backward compatibility (ie remove old sets that are still used in supported past releases). Moreover, I'm probably missing something but how does a recent CMSSW version know that it should look at V00-00-01 and not at V00-00-00? Both are found under CMSSW_DATA_PATH. In any case, Slava Kl has prepared the last two map sets much earlier than expected, and I would like to add these as well, will this result in a V00-00-02 and yet another copy of all the files, or can we update (or drop) V00-00-01 given it did not go to an actual release yet? |
On 6/23/17 2:29 PM, Nicola Amapane wrote:
Yes but in this case V00-00-01 contains exactly the same files of
V00-00-00, plus two sets in addition which do not interfere with
supported past releases. There is really no point in duplicating these
48k files, couldn't we have added the new files and created a new set
only when we really break backward compatibility (ie remove old sets
that are still used in supported past releases).
Moreover, I'm probably missing something but how does a recent CMSSW
version know that it should look at V00-00-01 and not at V00-00-00? Both
are found under CMSSW_DATA_PATH.
Applications use FileInPath and look for relevant package data in
$CMSSW_SEARCH_PATH
which contains, in particular (expanded)
$CMSSW_RELEASE_BASE/external/$SCRAM_ARCH/data
so they will find MagneticField/Interpolation/ in
$CMSSW_RELEASE_BASE/external/$SCRAM_ARCH/data/MagneticField/Interpolation/data
applications do not look in $CMSSW_DATA_PATH
In any case, Slava Kl has prepared the last two map sets much earlier
than expected, and I would like to add these as well, will this result
in a V00-00-02 and yet another copy of all the files, or can we update
(or drop) V00-00-01 given it did not go to an actual release yet?
A new version will correspond to another copy.
There is no practical way to avoid copies at this level:
basic distribution pattern that breaks the idea of cleanup in old area
is when
a new version of the data external has the same file names but different
content of the files.
For the pattern with the new package always adding new files and not
modifying the old ones
it is indeed more practical to overwrite old locations,
but technically this would require re-release and re-installation of old
releases.
…
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#1 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AEdcbg3oTsNlIoPjMarDueH-J2irMiaOks5sHC4vgaJpZM4Ny4ks>.
|
the incremental solution would be to instead distribute every field map version subdirectory as a separate external |
What do you prefer? I am now ready to make a PR for the the last two
table folders of this map set. If you are not concerned about this
duplication I would go on in the usual way.
…On 24-Jun-17 1:13, Slava Krutelyov wrote:
the incremental solution would be to instead distribute every field map
version subdirectory as a separate external
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#1 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AE7eUEO9rmimuo9OhRY9HrYv9K4dse_pks5sHEakgaJpZM4Ny4ks>.
|
It would be good to get some feedback here from David and Shahzad. |
Is there another iteration after this one expected? (Indeed I thought the last iteration was the last ...)
Cheers,
David
On 27 Jun 2017, at 20:10, Slava Krutelyov <notifications@github.com<mailto:notifications@github.com>> wrote:
It would be good to get some feedback here from David and Shahzad.
@davidlange6<https://github.com/davidlange6> @smuzaffar<https://github.com/smuzaffar>
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub<#1 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AEzywwC1x6MgZCnWqdUHctS3KiXaXjhMks5sIUVngaJpZM4Ny4ks>.
|
@namapane, as @slava77 mentioned, if we want to avoid duplication then we have to bundle these maps in separate externals which means you have to update cmssw code too. How much more data you would like to include ? Note that data-MagneticField-Interpolation github repo is nearly 500MB now and there is 1GB size limitation on github repos. |
@davidlange6, These would be the last two sets of tables for this map
set. (I actually did not expect them to be ready in time for 93X and
that's why I made a PR for the first two)
@smuzaffar: the additional data is about 170 MB, so we are not hitting
the quota anytime soon. Also, once these maps will become the default
(hopefully in 93X) it will be possible to remove the following sets
which will be by then obsolete:
grid_1103l_071212_3t
grid_1103l_071212_3_5t
grid_120812_3_8t_v7_large
grid_120812_3_8t_v7_small
grid_130503_3_5t_v9_large
grid_130503_3_8t_v9_large
grid_130503_3_8t_v9_small
(though I don't know if removal really frees quota given that files
remain in the github history?)
BTW, in fact one of these maps sets (grid_120812_3_8t_v7_small) can be
removed right away as it has been superseded before being adopted in any
release. Any objections in removing it?
Anyhow, I have a mild preference not to separate this in different
externals for the time being, unless you have strong opinions.
Cheers,
Nicola
…On 27-Jun-17 21:27, David Lange wrote:
Is there another iteration after this one expected? (Indeed I thought
the last iteration was the last ...)
Cheers,
David
On 27 Jun 2017, at 20:10, Slava Krutelyov
***@***.******@***.***>> wrote:
It would be good to get some feedback here from David and Shahzad.
@davidlange6<https://github.com/davidlange6>
@smuzaffar<https://github.com/smuzaffar>
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on
GitHub<#1 (comment)>,
or mute the
thread<https://github.com/notifications/unsubscribe-auth/AEzywwC1x6MgZCnWqdUHctS3KiXaXjhMks5sIUVngaJpZM4Ny4ks>.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#1 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AE7eUJ_M17F76pjDAd29DJxM2fjfNZWeks5sIVeNgaJpZM4Ny4ks>.
|
I am fine with keeping these in single external. |
For future field maps. No change in existing functionality or current default behavior.
cf: https://indico.cern.ch/event/642366/contributions/2606389/