Project

General

Profile

Feature #19877

Add a scaling factor for all glideins limits in the entries

Added by Marco Mambelli over 1 year ago. Updated 8 months ago.

Status:
Closed
Priority:
Normal
Category:
-
Target version:
Start date:
05/03/2018
Due date:
% Done:

0%

Estimated time:
Stakeholders:

Factory Ops

Duration:

Description

Entries have per entry and per frontend limits, that limit the number of glideins subbmitted, idle, ... at a specific entry.
Some organizations (CMS, OSG, ... ) may have more than one factory submitting to the same entry, though they want to enforce a global limit across all the factories they manage.
Traditionally operators have been adjusting the numbers by hand (3 factories -> all limits are divided by 3) but this brings a lot of numbers to change if the numbers of factories changes.
Would be useful to have a scaling factor "num_fcatories" for all these numbers.
num_factories will be in the factory configuration (glideinWMS.xml) will have a default of 1, can be defined at the global level in the configuration file, can be defined also in the entry-group level, overriding the default or global value, and in the entry level, overriding the default or global or entry-group value.
All limits in "<default_per_frontend..." and " <per_entry ..." will be scaled down dividing them by the num_fcatories value for that entry_group or entry.

Here from Jeff's emails formulating the request

I propose a simple solution to this that at least makes reconfiguring less painful. Can we please have a new attribute for entries that states num_factories or something along those lines? Then for the JINR case, we would have set what the site wanted total, 400 glideins. And then factory logic should take this number and divide by num_factories=4 before applying the real limit.

When factories come and go we could update the number in the entry from num_factories="4" to num_factories="3", etc. This is much better than having to find an old ticket where the admin stated their site totals, and recalculate by hand.

Also there should be a global default that all entries accept but then we should be able to override it at the entry level.

...

Let's think of a simple example. Say for a site we have a single entry and the admins says, "no more than 100 glideins at this CE" but we're configuring that entry to be in both SDSC and CERN.

I want to be able to set:
               <default_per_frontend glideins="100" held="20" idle="20"/>
               <per_entry glideins="100" held="20" idle="20"/>

and a new atribute, num_factories=2

rather than stopping submission when they hit any of those limits for each factory, the real factory logic stops as if it were configured like:
               <default_per_frontend glideins="50" held="10" idle="10"/>
               <per_entry glideins="50" held="10" idle="10"/>

So the entry for CERN and SDSC factory will each stop once they are running 50 running pilots. the total pilots running at the site would be 50+50 = 100, which is exactly the limit the admins told us to set the cap the site at.


Related issues

Related to GlideinWMS - Feature #19946: Factory Operations suggestions summaryNew05/14/2018

History

#1 Updated by Marco Mascheroni over 1 year ago

  • Assignee set to Marco Mascheroni

#2 Updated by Marco Mascheroni about 1 year ago

  • Status changed from New to Work in progress

#3 Updated by Marco Mascheroni about 1 year ago

  • Related to Feature #19946: Factory Operations suggestions summary added

#4 Updated by Marco Mascheroni about 1 year ago

  • Target version changed from v3_4_1 to v3_5

#5 Updated by Marco Mascheroni about 1 year ago

  • Status changed from Work in progress to Feedback
  • Assignee changed from Marco Mascheroni to Marco Mambelli

#6 Updated by Marco Mambelli 11 months ago

  • Assignee changed from Marco Mambelli to Marco Mascheroni

#7 Updated by Marco Mambelli 9 months ago

  • Target version changed from v3_5 to v3_4_3

#8 Updated by Marco Mascheroni 9 months ago

  • Status changed from Feedback to Resolved

#9 Updated by Marco Mascheroni 9 months ago

  • Assignee changed from Marco Mascheroni to Marco Mambelli

#10 Updated by Marco Mascheroni 9 months ago

  • Assignee changed from Marco Mambelli to Lorena Lobato Pardavila

#11 Updated by Marco Mascheroni 9 months ago

Review the documentation of this as well, please.

#12 Updated by Marco Mambelli 8 months ago

  • Status changed from Resolved to Closed


Also available in: Atom PDF