HTCondor job submission tags

Question

I want to run different batches of jobs on our HTCondor pool. Let's say 10 jobs of Type1, 20 jobs of Type2 and so on. Each of these job types should get new jobs when the current jobs are finished.

With just one type I use a simply query if all jobs are finished or if the time limit for the whole job batch passed. If one of these requirements is fulfilled the next iteration of x jobs is submitted to the cluster.

This is done by a small function (written in Lua, which is not really important for the question):

function WaitForSims(CheckupDelay)
    while io.popen([[condor_q -format "%d\n" clusterid]]):read('*all'):len()~=0 do
        os.execute("echo Checkup timestamp: "..os.date("%x %X"))
        os.execute(string.format("timeout %d 1>nul",CheckupDelay))
    end
end

Is there a possibility to separate the jobs of Type1, Type2 and Type3 and check them independently? Currently it checks for all jobs as my current user.

Adding a tag or something to the jobs would be ideal, as I could simply change the checkup call. In the documentation I couldn't find anything which is easy to add, I could remember the JobID-s, but then I'll have to store those adding more complexity.

Answer 1

Linked Answer

Solution could be found in another answer, I didn't find where it is described in the documentation though.

In the job.sub file add:

+YourCustomVarName = 1
+YourCustomStringName = "String"

For checking against it use:

condor_q -constraint 'YourCustomVarName == 1' -f "%s" JobStatus

or

condor_q -constraint "YourCustomStringName == \"String\"" -f "%s" JobStatus

(handling of quotations could vary)