HTCUpgrade: Difference between revisions
No edit summary |
|||
Line 28: | Line 28: | ||
= How to keep access to your jobs = | = How to keep access to your jobs = | ||
As the 2 HTCondor versions will be live in parallel, you will still be able to manage your jobs on the old HTC version from the | As the 2 HTCondor versions will be live in parallel, you will still be able to manage your jobs on the old HTC version from the upgraded mX machines (and vice-versa). | ||
Revision as of 11:21, 4 July 2022
Overview
We plan to upgrade HTCondor from version 8.9 to 9.0.
The main benefits of the new version is to be compatible with tokens, the replacement of certificates for GRID authentification.
As it is a major change, it will require a complete reinstallation of all services.
But this WILL NOT impact your jobs, as compute resources are drained only when empty. You will be able to access jobs on both cluster versions.
Migration Schedule
As of 04/07, already 70% of job slots have been migrated.
Most of GRID jobs are already using the new version.
For you to have a more transparent experience during this upgrade, please always use mshort.iihe.ac.be or mlong.iihe.ac.be to connect to the cluster.
- Monday 8PM
- m1 will be excluded from mshort
- m4/m5 will be excluded from mlong
- Tuesday 8PM
- m1/m4/m5 will not be accessible anymore and will be reinstalled
- Wednesday 12AM
- m2/m3 will be excluded from mshort
- m6/m7 will be excluded from mshort
- mshort will lead to upgraded HTC on m1
- mlong will lead to upgraded HTC on m4/m5
- Wednesday 8PM
- m2/m3/m6/m7 will not be accessible anymore and will be reinstalled
How to keep access to your jobs
As the 2 HTCondor versions will be live in parallel, you will still be able to manage your jobs on the old HTC version from the upgraded mX machines (and vice-versa).
- Old HTC version:
condor_q <==> condor_q -n schedd01 <==> condor_q -pool testumd-htcmaster.wn.iihe.ac.be -name schedd01
- New HTC version:
condor_q <==> condor_q -n schedd03 <==> condor_q -pool cm.wn.iihe.ac.be -n schedd03
So once the mX machines are upgraded, you will be able to access the old HTC verion cluster with:
condor_q -pooltestumd-htcmaster.wn.iihe.ac.be -name schedd01
Or on the contrary, directly start submitting on the new HTC version cluster from the yet-to-be-upgrade mX machines (ie you can do it right now) with:
condor_submit -pool cm.wn.iihe.ac.be -n schedd03 MYFILE.SUB condor_q -pool cm.wn.iihe.ac.be -n schedd03