double the reserve for on-demand strip cluster from raw

[slava77]

This is a cheap temporary solution (about 2.7 MB cost per stream).
It resolves several crashes reported in HLT step of 810pre7 relvals
https://hypernews.cern.ch/HyperNews/CMS/get/relval/5231.html

The problematic "Run 274199, Event 268225325, LumiSection 137" had a total of 132,303 clusters to be unpacked, while the previous limit was 132,303.

A configurable and possibly more graceful solution is expected with
https://its.cern.ch/jira/browse/CMSTRACK-145

@fwyzard @VinInn @venturia

[cmsbuild]

A new Pull Request was created by @slava77 (Slava Krutelyov) for CMSSW_8_1_X.

It involves the following packages:

RecoLocalTracker/SiStripClusterizer

@cmsbuild, @cvuosalo, @slava77, @davidlange6 can you please review it and eventually sign? Thanks.
@makortel, @forthommel, @yduhm, @GiacomoSguazzoni, @gbenelli, @rovere, @VinInn, @nickmccoll, @jlagram, @gpetruc, @OlivierBondu, @threus this is something you requested to watch as well.
@slava77, @Degano, @smuzaffar you are the release manager for this.

cms-bot commands are list here #13028

[slava77]

@cmsbuild please test

[cmsbuild]

The tests are being triggered in jenkins.
https://cmssdt.cern.ch/jenkins/job/ib-any-integration/13627/console

[VinInn]

if critical to HLT, shall not be implemented in 8_0_X (patch) as well?

[slava77]

I'll make a PR for 80X once this goes in.
This is not critical for actual data operations, if I understand correctly, this error never happened (or at least was not reported) in 80X data taking.
Similar to the past experience with this reserve, I'm assuming that this is mainly an issue with relvals which may have no prescales applied.

[slava77]

@fwyzard @Martin-Grunewald
if I make a backport PR for 80X, do you need it in 8_0_10_patchX as well?

[cmsbuild]

+1
Tested at: 71d0571
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-14953/13627/summary.html

[slava77]

@cmsbuild please test
comparisons failed for the last attempt

[cmsbuild]

The tests are being triggered in jenkins.
https://cmssdt.cern.ch/jenkins/job/ib-any-integration/13631/console

[cmsbuild]

+1
Tested at: 71d0571
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-14953/13631/summary.html

[cmsbuild]

Comparison is ready
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-14953/13631/summary.html

[slava77]

+1

for #14953 71d0571

• jenkins tests pass and comparisons show no differences

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next CMSSW_8_1_X IBs (tests are also fine). This pull request requires discussion in the ORP meeting before it's merged. @slava77, @davidlange6, @Degano, @smuzaffar

[Martin-Grunewald]

@slava77
Yes, please backport to 80X.
Thanks!

[slava77]

On 6/22/16 9:49 PM, Martin Grunewald wrote:

@slava77 https://github.com/slava77
Yes, please backport to 80X.
Thanks!

only to 80x (8_0_12++) or also an older patch release?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#14953 (comment), or
mute the thread
https://github.com/notifications/unsubscribe/AEdcbq1ImPa0Ecg5AtrvEWyO1kFXnOwyks5qOhBvgaJpZM4I7_iK.

\

[Martin-Grunewald]

@slava77
Only the normal 80X

[davidlange6]

+1

[fwyzard]

Thank you guys, I like being ignored.

Can you please fix the underlying container so that it does not allow increasing its size, instead of increasing the initial reserve and hope that we do not go over it ?

[fwyzard]

Can you please add a protection to the code that uses this, so that it does not try to store more objects than the container can hold ?

[slava77]

On 6/24/16 2:09 AM, Andrea Bocci wrote:

Thank you guys, I like being ignored.

Can you please fix the underlying container so that it does not allow
increasing its size, instead of increasing the initial reserve and hope
that we do not go over it ?

Hi Andrea,

you are not ignored,
the long term solution will be made as a part of
https://its.cern.ch/jira/browse/CMSTRACK-145
Please add your suggestions there.

Slava

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#14953 (comment), or
mute the thread
https://github.com/notifications/unsubscribe/AEdcbmnY5dxAeQxjQVrzhVwbdm_SEcR5ks5qO57bgaJpZM4I7_iK.

[slava77]

On 6/24/16 2:10 AM, Andrea Bocci wrote:

Can you please add a protection to the code that uses this, so that it
does not try to store more objects than the container can hold ?

I don't quite understand this.
Client code is not supposed to know the capacity of the on-demand
containers.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#14953 (comment), or
mute the thread
https://github.com/notifications/unsubscribe/AEdcbvoGZLVvVq7lYmhwijOO96Zz_AFAks5qO58cgaJpZM4I7_iK.

[fwyzard]

I don't quite understand this. Client code is not supposed to know the capacity of the on-demand containers.

In principle I agree with you - however here the problem is that the "on demand container" exposes an interface similar to that of a vector, but it will throw an exception or crash if the client actually tries to resize it after it has been filled, by pushing more elements that the reserved capacity is.