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
make a production-like workflow for HI 2018 #24587
Comments
A new Issue was created by @slava77 Slava Krutelyov. @davidlange6, @Dr15Jones, @smuzaffar, @fabiocos, @kpedro88 can you please review it and eventually sign/assign? Thanks. cms-bot commands are listed here |
assign pdmv |
@slava77 , @mandrenguyen this is due to the absence of pfNoPileUpJME, exactly as @slava77 pointed out. Should I update the existing pull request or create a new one? |
@stepobr The original PR is merged already, so you'll need a new one. At some point we discussed setting a threshold on ak8 jets such that it doesn't use up any CPU (since we don't use it anyway). Perhaps that can be done at the same time. But the priority is get something that runs. |
Hi - we will discuss and finalize how we address this in tomorrow's meeting, but just for record keeping here: my inclination is to not re-add this collection as it is only used for jet collections useless in HI that we keep only to not potentially break things in reco down the line. Unless this collection is already dummy, we should just feed a filtered particleFlow with pt cut of 9999. to cut timing or feed particleFlow directly and kick up the the inputEtMin to 9999. on the FastJetProducer itself. The latter would reduce size completely and cuts timing for each algo by ~3. If it is dummy it would operate like former. This is only from what I understand - if I am overlooking something please let me know. |
it looks like this missing pfNoPileUpJME is the only thing broken in the "production" setup using |
@slava77 @mandrenguyen is there a follow-up of this issue? |
In order to avoid future possible problems as we had in T0 replay recently, it would be still nice to resolve this issue and make a matrix workflow. It looks like a bit more complete set in the DQM step would be to have |
@zhenhu @prebello I have discussed with @icali, it would be good to have a realistic chain that combines all the ingredients:
|
We collected the run 325174 this morning that has the (quasi) final L1+HLT menu. It includes both the datasets with full and reduced format. |
I tried to add two new workflows with HIMinimumBias0 or HIMinimumBiasReducedFormat0 as input, The workflows run without crashes, but give me some warning messages, such as:
Above errors can by reproduced by running the two new workflows:
The main cmsDriver used is:
Do you have any suggestion how to proceed? |
I saw the same error using the standard GT provided by the cmsDriver command. However, changing the GT to 103X_dataRun2_Prompt_v2, it solved the issue. However, I'm not sure what would it be the best GT to be used. |
Hi @icali , the new GT does not work for me. To reproduce my error, you can first add the two new wfs as in e0e3a60
|
Hi all, By comparing @icali 's config with mine, we finally found the source of the errors. In Ivan's config, he used:
, while I was using:
The reason of the above difference is because I have So, in order to make my reco step work, I committed a new version which is exactly same as Ivan's config: But the EI step still has errors. Any suggestion on this?
|
On 10/25/18 2:35 AM, Zhen Hu wrote:
But the EI step still has errors. Any suggestion on this?
this is expected now
and should be fixed by #24958
|
#24958 has been merged (as well as its 10_3_X backport), so in next IB you should not see any more the issue |
Looking at 104X IB 2018-11-07:
I agree. This satisfies the original request. I'm closing this issue. |
HI data taking for 2018 is supposed to be written in AOD and to be running without miniAOD outputs.
For some time I was using
-s RAW2DIGI,L1Reco,RECO,EI,DQM:@standardDQM
with output directed to AOD[SIM]. This stopped working in 10_3_0_pre3 due to some updates from #24380 .I think that we need a relval setup to test the desired production configuration.
@mandrenguyen please comment [or redirect] on the needed processing steps in production.
Should we be running miniAOD, even if it's not needed?
The text was updated successfully, but these errors were encountered: