Re: [HTCondor-users] how to get current 'fair share' quota/weight of a group?

[Iain Bradford Steers <iain.steers@xxxxxxx>]

Hi Thomas,

Which version of condor are you running on your central manager?

In the 8.5 series the negotiator now sends it's accounting information to the collector as an ad per item in the tree.

Alternatively using the python bindings you can instantiate a Negotiator object and use .getPriorities().

The Collector is obviously much faster.

Cheers, Iain

You can see the list with condor_status  -any -const 'MyType =?= "Accounting"'
For example:

Parts of the tree also have a couple of others like ConfigQuota set.

UnchargedTime = -6
LastUsageTime = 1474901308
AuthenticatedIdentity = "condor@fsauth"
IsAccountingGroup = false
PriorityFactor = 100000.0
MyType = "Accounting"
AccountingGroup = "group_u_X"
Name = "group_u_x"
WeightedAccumulatedUsage = 128064692224.0
LastUpdate = 1474901308
WeightedResourcesUsed = 7677.0
ResourcesUsed = 7677
TargetType = "none"
AccumulatedUsage = 127910363136.0
LastHeardFrom = 1474901320
BeginUsageTime = 1447772083
DaemonStartTime = 1473746649
UpdateSequenceNumber = 0
Priority = 937499392.0
WeightedUnchargedTime = -6.0
________________________________________
From: HTCondor-users [htcondor-users-bounces@xxxxxxxxxxx] on behalf of Thomas Hartmann [thomas.hartmann@xxxxxxx]
Sent: 26 September 2016 16:45
To: HTCondor-Users Mail List
Subject: [HTCondor-users] how to get current 'fair share' quota/weight of a     group?

Hi all,

on a user's question [1], I am currently looking for a ClassAd with the
current weighting/fair share quota of a given group?

I.e.,
we have several dynamic quota groups. If I remember correctly, by
default the resource usage over the last three days is taken into
account for weighting quotas.
So, I have been skimming, so far unsuccessfully, through
  -state
  -collector
looking for an ad with a flat value of the current quota.
But is there actually such a 'scalar' ad?

If not, I assume, that the quota is calculated by the collector each
negotiation cycle anew for any queued job with its resource requests
(including ConcurrencyLimits, ChildAccountingGroup,...)?
And if so, is there an 'easy' algorithm to calculate oneself the
collector's hypothetical current quota for a job against everybody else
being queued?

Cheers and thanks,
  Thomas


[1]
who would like to try to 'safe up' its fair share quota for some more
higher priority work load - which is from my point of view really *not*
a good approach, i.e., folding some oscillating behaviour into the
system,...