Re: [HTCondor-users] Job NOT considered by the matchmaker

[Alessandro Italiano <alessandro.italiano@xxxxxxxxxx>]

Hi

since Saturday âRequested Resourcesâ is working as aspected. 

-  âRequested Resourcesâ >= âRes InUse"

[root@htc-ctl-01 ~]# condor_userprio 

Last Priority Update: 11/30 09:17

Group                            Config     Use    Effective   Priority   Res   Total Usage  Time Since Requested 

  User Name                       Quota   Surplus   Priority    Factor   In Use (wghted-hrs) Last Usage Resources 

------------------------------- --------- ------- ------------ --------- ------ ------------ ---------- ----------

group_cms.grid                       0.50 ByQuota                1000.00   3784    767560.12      <now>       4696 

 How the Negotiator/condor_userprio obtains the underline data in order to calculate the

   âallocationâ/ âRequested Resourcesâ value

the following entry reports +/- the same value

condor_status  -any -const 'MyType =?= "Accounting" && Name=="group_cms.grid"' -l

In the file matchmaker.cpp I was not able to find the answer.

Ale

On 27 Nov 2020, at 11:17, Alessandro Italiano <alessandro.italiano@xxxxxxxxxx> wrote:

Hi

even with the last htcondor version, 8.8.12, I continue to experience this odd behaviour.

Jobs belonging to AccountingGroup group_cms.grid are not considered for Negotiation

Here [1] you can find Negotiator cycle log at D_FULLEDEBUG

[root@htc-ctl-01 ~]# condor_userprio 

Last Priority Update: 11/27 11:10

Group                              Config     Use    Effective   Priority   Res   Total Usage  Time Since Requested 

  User Name                         Quota   Surplus   Priority    Factor   In Use (wghted-hrs) Last Usage Resources 

--------------------------------- --------- ------- ------------ --------- ------ ------------ ---------- ----------

group_cms.grid                         0.90 ByQuota                1000.00   1552    535551.81      <now>       1207

  cms004@ReCaSCluster                                   16200.19   1000.00     24      1224.63      <now>

  cms016@ReCaSCluster                                 1614032.00   1000.00   1528    516462.25      <now

[root@htc-ctl-01 ~]# condor_q -global -cons '(x509UserProxyVOName == "cms")'| grep "Total for query" 

Total for query: 128 jobs; 0 completed, 25 removed, 40 idle, 63 running, 0 held, 0 suspended 

Total for query: 134 jobs; 0 completed, 21 removed, 40 idle, 73 running, 0 held, 0 suspended 

Total for query: 106 jobs; 0 completed, 10 removed, 40 idle, 56 running, 0 held, 0 suspended 

11/27/20 11:14:46 group quotas: fairshare (2): group= group_cms  quota= 3451.21  allocated= 0  requested= 0

11/27/20 11:14:46 group quotas: fairshare (1): group= group_cms.grid  quota= 3106.09  requested= 1230

11/27/20 11:14:46 group quotas: fairshare (2): group= group_cms.grid  quota= 3106.09  allocated= 1230  requested= 0

11/27/20 11:14:46 Group group_cms.grid - skipping, at or over quota (quota=3106.09) (usage=1536) (allocation=1230)

Ale

[1]

https://istnazfisnucl-my.sharepoint.com/:u:/g/personal/ale_infn_it/EYnHrCJ15-9HgMAMsk1unwMBBhAQVySYdUp_macs0N_y3w?e=g03mKy

On 24 Nov 2020, at 08:15, Alessandro Italiano <alessandro.italiano@xxxxxxxxxx> wrote:

Hi

thanks for the answer.

the cluster is running condor 8.6.13. 

We have completed the migration from CREAM to CONDOR-CE, so we have 3 SCHEDDs at 8.8.10

After a long debug session, I have[I believe] evidence the issue is related with the SLOT_WEIGHT.

It seems multicore jobs submitted through the condor-ce are not evaluated in a weighted manner by the Negotiator.

We have scheduled a HTcondor upgrade