Project

General

Profile

Idea #24569

Some requests/suggestions form HEPCloud

Added by Marco Mambelli 4 months ago. Updated 17 days ago.

Status:
New
Priority:
High
Category:
-
Target version:
Start date:
07/08/2020
Due date:
% Done:

0%

Estimated time:
(Total: 0.00 h)
Stakeholders:
Duration:

Description

Here some requests from a discussion w/ Steve Timm

These are incomplete/confusing notes for internal use. Tickets will be generated

the kill or no kill should be configurable
Glidein should stay there also if no requests
Credential management unique in the system (unique key per credential, referenceable in different places)

calculating glideins and advertising should be decoupled

exposing the advertising:
- produces and advertises is a string list of names
A lot of redindant code because of that
e.g. give me what said in the config

Being able to debug: why is the frontend requesting one glidein instead of 200

Extend FOM to multiple factories

Repeat the same Factory configuration TotalGlideinsPerEntry and TotalGlideinsPerEntry
-> LimitPreDEPer Entry brought internally
Tweak the constraint for the idle jobs
DE is not using the match expression
Condor status is not returning Idle running and idle busy
(elimeinate partitionable and older than a certain glidein to retire)

There are idle jobs in the queue, no glidein that can run them
(e.g. jobs at NERSC have not enough time in the queue)

A 20 node job ends up having only one glidein reporting back, the other died.

Project ID differs, 
there are 5 
tag or expression for site-specific values key-value, w/ default value otherwise

Better credential handling, more credentials, attributes connected

Send the pointer to use parameter in submit file.

Existing job classification in the DE
is inadequate, if_then


Subtasks

Bug #24610: Glidein pressure too low for HPC systems in HEPClodNewMarco Mambelli

History

#1 Updated by Marco Mambelli 3 months ago

  • Start date changed from 06/26/2020 to 07/08/2020
  • Due date set to 07/08/2020

due to changes in a related task: #24610

#2 Updated by Marco Mambelli about 1 month ago

  • Target version changed from v3_6_4 to v3_6_5

#3 Updated by Marco Mambelli 17 days ago

  • Target version changed from v3_6_5 to v3_6_6


Also available in: Atom PDF