HTCondor

Introduction

At PIC HTCondor is the current production batch system replacing the old Torque/Maui environment.

The aim of this document is to show how to submit jobs to the HTCondor infrastructure to all of the non-grid users of the PIC batch system. In other words, this document is a guide to submit local jobs to the HTCondor infrastructure at PIC.

We strongly recommend to look at the HTCondor User Manual [1] if you want a deeper approach to the HTCondor concepts.

This User Guide begins with a presentation of the batch system concepts in HTCondor in comparison with the old Torque/Maui ones. Then, there is a Quick Start section focused on the minimum knowledge needed to submit a job in an HTCondor pool. The remaining sections try to give a deeper approach in how to submit, monitor and remove jobs in PIC HTCondor pool.

Furthermore, you can access to the HTCondor User Guide slides and a git repository with several examples.

• HTcondor User Guide Tutorial: https://docs.google.com/presentation/d/1-64fEcfLyxLzSpZH-tbV-SjMAc0CvMpWhE0I4oaDKkQ/edit?usp=sharing
• HTCondor Tutorial examples: https://github.com/PortdInformacioCientifica/htcondor-tutorial

Basic batch concepts

HTCondor does not work as other batch systems where you submit your job to a differentiated queue that has some specifications. It employs the language of ClassAds (the same concept of classified advertisements) in order to match workload requests and resources. In other words, the jobs and the machines have their particular attributes (number of CPUs, memory, etc.) and the central manager of HTcondor does the matchmaking between these attributes.

Furthermore, similarly to Torque/Maui, there is the concept of fair-share which aims at ensuring that all groups and users are provided resources as needed in correspondence to their respective quota (e.g. the Atlas T2 quota equals to 9% of our resources). The fair-share concept implies that your jobs and the jobs of your experiment will have a greater priority while they are agreed at or below the share, if you are consuming more resources than your share, then the next job with more priority should belong to another experiment.

When a user submits a job from the user interface, it is queued in the HTCondor queue system (schedd) and, according to its priority and its requirements, the job is assigned by the batch system's Central Manager (running collector and negotiator daemons) to be executed in a Worker Node (startd) that matches its requirements. Once the job has finished, files such as the job log, the standard output and the standard error are retrieved back from the Worker Node to the submit machine.

Quick start

Before taking a deeper view in all the elements of the job submission, we will show you the basic commands for a quick start guide to HTcondor.

In our old Torque/Maui environment, the user would log into a machine, prepare the input and submit jobs to a queue using qsub command. Now, in a very similar way, the user logs into the User Interface, prepares a submit file, and then creates and inserts jobs into the queue using a condor_submit command.

Since November 2019, all the users can submit directly their jobs from the user interfaces.

Next, you can find the basic skeleton of an HTcondor submit file (called test.sub in this case).

$ cat test.sub
executable = test.sh
args = -c 1 -t 60
output = condor.out
error = condor.err
log = condor.log

queue

The script just executes a stress command (a simple workload generator for UNIX systems). You can specify the number of cpus you want to stress (-c 1) and the timeout of the process (-t 60 seconds).

$ cat test.sh
#!/bin/bash

/bin/stress $@

This example can be easily understood as follows: it must include the executable with your script or command, the arguments (args) of your command and where to store the STDOUT (output), the STDERR (error) and the HTCondor log which reports the status of the job. Finally, you can find the "queue" command which tells condor how many instances of the job you want to run ("queue 1", or simply "queue", to submit one, or "queue N" to submit N jobs, for instance). You can find more information about these variables in the next sections of this manual.

Then, you can submit your job using condor_submit.

$ condor_submit test.sub 
Submitting job(s).
1 job(s) submitted to cluster 281.

Make sure that your script or executable is correctly created, for example, this means that it has the correct execution permissions and your script has the shebang (the character sequence line which starts with #! at the beginning of the script). In other words, your executable has to be runnable from the command line.

On the other hand, in order to monitor the status of your job, you can query the queue with the condor_q command (in a similar way as you do with qstat in Torque). The default condor_q output can vary slightly from one version to another in HTCondor.

$ condor_q


-- Schedd: submit01.pic.es : <193.109.174.82:9618?... @ 02/25/19 13:07:52
OWNER   BATCH_NAME    SUBMITTED   DONE   RUN    IDLE  TOTAL JOB_IDS
cacosta ID: 281      2/25 13:07      _      _      1      1 281.0

Total for query: 1 jobs; 0 completed, 0 removed, 1 idle, 0 running, 0 held, 0 suspended
Total for cacosta: 1 jobs; 0 completed, 0 removed, 1 idle, 0 running, 0 held, 0 suspended  
Total for all users: 1 jobs; 0 completed, 0 removed, 1 idle, 0 running, 0 held, 0 suspended

It returns the owner, the batch name of your job, the submission date, the status (Done, Run or Idle) and the JobIds.

Using the option -nobatch reports an output that does not group the jobs.

$ condor_q -nobatch


-- Schedd: submit01.pic.es : <193.109.174.82:9618?... @ 02/25/19 13:10:10
 ID      OWNER            SUBMITTED     RUN_TIME ST PRI SIZE CMD
 281.0   cacosta         2/25 13:09   0+00:00:00 I  0    0.0 test.sh -c 1 -t 60

Total for query: 1 jobs; 0 completed, 0 removed, 1 idle, 0 running, 0 held, 0 suspended
Total for cacosta: 1 jobs; 0 completed, 0 removed, 1 idle, 0 running, 0 held, 0 suspended 
Total for all users: 1 jobs; 0 completed, 0 removed, 1 idle, 0 running, 0 held, 0 suspended

Finally, to remove the jobs, you can use condor_rm that works as qdel in Torque/Maui.

$ condor_rm 281
All jobs in cluster 281 have been marked for removal

Submitting your jobs

After a basic view of how to submit a job, we are going to explain more details about the job submission, in particular about the options of the submit file. You can find detailed information about condor_submit and the characteristics of the submit file in the documentation [2].

First of all, it has to be mentioned that the only options that are mandatory in a submit file are the "executable" and the "queue" command.

Executable, input, arguments, outputs, errors and logs

You must specify your executable and you can specify the input, output and error logs in your submit files as we have seen before:

executable = exec
input  = input.txt
output = out.txt
error  = err.txt
log    = log.txt

Thus, you can specify the location of the input of your application, considering that HTCondor uses the input to pipe into the stdin of the executable. On the other hand, there is the output which contains the standard output (stdout) and the error which contains the standard error (stderr). The log file reports the status of the job by HTCondor. If any of these options is not defined in the submit file, HTCondor redirects standard output and standard error to /dev/null.

In the log file, you can see the submission host, the node where the job is executed, information about the memory consumption and a final summary of the resources used by your job. Here you have an example of a typical log file:

$ cat test-736.0.log 
000 (736.000.000) 03/25 10:00:22 Job submitted from host: <193.109.174.82:9618?addrs=193.109.174.82-9618&noUDP&sock=961738_da40_3>
...
001 (736.000.000) 03/25 10:00:40 Job executing on host: <192.168.101.48:9618?addrs=192.168.101.48-9618&noUDP&sock=14755_47fd_3>
...
006 (736.000.000) 03/25 10:00:49 Image size of job updated: 972
	1  -  MemoryUsage of job (MB)
	952  -  ResidentSetSize of job (KB)
...
005 (736.000.000) 03/25 10:00:50 Job terminated.
	(1) Normal termination (return value 0)
		Usr 0 00:00:10, Sys 0 00:00:00  -  Run Remote Usage
		Usr 0 00:00:00, Sys 0 00:00:00  -  Run Local Usage
		Usr 0 00:00:10, Sys 0 00:00:00  -  Total Remote Usage
		Usr 0 00:00:00, Sys 0 00:00:00  -  Total Local Usage
	111  -  Run Bytes Sent By Job
	28  -  Run Bytes Received By Job
	111  -  Total Bytes Sent By Job
	28  -  Total Bytes Received By Job
	Partitionable Resources :    Usage  Request Allocated 
	   Cpus                 :     0.02        1         1 
	   Disk (KB)            :    17           1    891237 
	   Memory (MB)          :     1         200       256 
...

It is worth mentioning that the log file allows us to monitor our jobs without doing several condor_q queries using the condor_wait command. You will find more information about condor_q and condor_wait later in this document.

$ condor_wait -status log-62.0.log 
62.0.0 submitted
62.0.0 executing on host <192.168.100.29:9618?addrs=192.168.100.29-9618+[--1]-9618&noUDP&sock=73457_f878_3>
62.0.0 completed
All jobs done.The command '''condor_wait''' is used to track the information in log file. This command will be explained in next sections.

ClusterId and JobId

When you submit multiple jobs, it is in general useful to assign unique filenames, for example typically containing the cluster and job ID variables (ClusterId and ProcId respectively). For instance:

executable = test.sh
input = input.txt
arguments = arg
output = test-$(ClusterId).$(ProcId).out
error = test-$(ClusterId).$(ProcId).err
log = test-$(ClusterId).$(ProcId).log

The job identifiers are $(ClusterId).$(ProcId) in HTcondor. The jobs in the queue are grouped in batches or clusters, the general number of your batch is the ClusterId while the different jobs inside your batch are defined by ProcId. In other words, if you submit only one job, you will obtain a $(ClusterId).0, while if you submit for instance 3 jobs using the same submit file, you will obtain $(ClusterId).0, $(ClusterId).1 and $(ClusterId).2. Furthermore, you can define a name for your batch using +JobBatchName option.

The next test.sub file submits two jobs to the queue.

executable = test.sh
arguments = -c 1 -t 60
output = test-$(ClusterId).$(ProcId).out
error = test-$(ClusterId).$(ProcId).err
log = test-$(ClusterId).$(ProcId).log

+JobBatchName="MyJobs"

queue 2

$ condor_submit test.sub 
Submitting job(s)....
2 job(s) submitted to cluster 740.

Using condor_q we can see the batch name and the jobs grouped.

$ condor_q 740

-- Schedd: submit01.pic.es : <193.109.174.82:9618?... @ 03/25/19 11:27:30
OWNER   BATCH_NAME    SUBMITTED   DONE   RUN    IDLE  TOTAL JOB_IDS
cacosta MyJobs       3/25 11:27      _      _      2      4 740.0-1

Total for query: 2 jobs; 0 completed, 0 removed, 2 idle, 0 running, 0 held, 0 suspended 
Total for all users: 24 jobs; 0 completed, 0 removed, 5 idle, 19 running, 0 held, 0 suspended

Using condor_q -nobatch we monitor the status of the jobs ungrouped.

$ condor_q -nobatch 740

-- Schedd: submit01.pic.es : <193.109.174.82:9618?... @ 03/25/19 11:27:33
 ID      OWNER            SUBMITTED     RUN_TIME ST PRI SIZE CMD
 740.0   cacosta         3/25 11:27   0+00:00:00 I  0    0.0 test.sh -c 1 -t 60
 740.1   cacosta         3/25 11:27   0+00:00:00 I  0    0.0 test.sh -c 1 -t 60

Total for query: 2 jobs; 0 completed, 0 removed, 2 idle, 0 running, 0 held, 0 suspended 
Total for all users: 24 jobs; 0 completed, 0 removed, 5 idle, 19 running, 0 held, 0 suspended

Requests

You can request the cpu, disk and memory that your jobs need. This is done by request_cpus, request_disk and request_memory options in your submit file. Take into account that you can use units in your requests.

executable = test.sh
args = -c 8 -t 60
output = test-$(ClusterId).$(ProcId).out
error = test-$(ClusterId).$(ProcId).err
log = test-$(ClusterId).$(ProcId).log

request_memory = 4 GB
request_cpus = 8
request_disk = 30 GB

queue

Thus, this job asks for 4 GB of RAM, 8 CPUs and 30 GB of disk.

$ condor_q 1462 -af RequestCpus RequestMemory RequestDisk
8 4096 31457280

Here you can see the report of the requested resources: the number of cpus, the memory in MB and the disk in KB.

There are default values already defined:

• 1 cpu
• 2 GB of memory per cpu
• 15 GB of disk per cpu

Take into account that the single-core and multi-core jobs are defined by the request_cpus option in your submit_file. The slot is created in the WNs with the resources that satisfy your request. Although you can ask for the number of slots you desire, note that our pool is better tuned for multicore jobs of 8 cpus (therefore it will be easier to satisfy such requests, hence these jobs will remain in queue shorter).

Note than in request_memory or request_disk, M or MB indicates MiB, G or GB indicates GiB, etc.

Flavours

The maximum walltime of your job can be specified using a flavour. There are 3 such flavours: short, medium and long.

• short: 3 hours
• medium: 48 hours
• long: 96 hours

executable = test.sh
args = -c 1 -t 60
output = output-$(ClusterId).$(ProcId).out
error = error-$(ClusterId).$(ProcId).err
log = log-$(ClusterId).$(ProcId).log

+flavour="short"

queue

If you do not choose any flavour explicitly, the flavour medium is the default which corresponds to 48 hours of walltime.

Once the job arrives at the time limit, it will be held and it remains in this status for 6 hours before being removed from the queue. Thus, the user can check the jobs held in the queue for 6 hours. You will find more information about the JobStatus in later sections.

The priority works in this order short > medium > long being the shortest jobs the ones with higher priority.

Moreover, the flavours strictly control the memory consumption of your jobs. Jobs that exceed 50% over the requested memory will be also held.

If for some reason your jobs need walltime over the 96 hours, please ask your contact.

Environment

The jobs find several grid variables defined, the $HOME variable and a general $PATH (/bin:/usr/local/bin:/usr/bin). However, the user may define environment variables for the job's environment by using the environment command.

For instance, for the next script and submission file:

$ cat test.sh 
#!/bin/bash

echo 'My HOME directory is: ' $HOME
echo 'My Workdir is: ' $PWD
echo 'My PATH is: ' $PATH
echo 'My SOFTWARE directory is: ' $SOFT_DIR

$ cat test.sub 
executable = test.sh
output = output-$(ClusterId).$(ProcId).out
error = error-$(ClusterId).$(ProcId).err
log = log-$(ClusterId).$(ProcId).log

queue

Once the job is executed, the next output file is obtained:

$ cat output-128.0.out 
My HOME directory is:  /nfs/pic.es/user/c/cacosta
My Workdir is:  /home/execute/dir_13143
My PATH is:  /bin:/usr/local/bin:/usr/bin
My SOFTWARE directory is: 

The $HOME directory is defined, there is not a $HOME/bin directory in the $PATH that is defined in the .bashrc and the $SOFT_DIR variable is also empty. Taking into account that these variables are known and defined, for instance, in the .bashrc, we can submit the job in different ways, using the environment command or directly adding the needed exports in your script.

• 1) Using environment command

$ cat test1.sub 
executable = test.sh
output = output-$(ClusterId).$(ProcId).out
error = error-$(ClusterId).$(ProcId).err
log = log-$(ClusterId).$(ProcId).log

environment=PATH=$ENV(PATH);SOFT_DIR=/software/dteam/

queue

$ cat output-129.0.out 
My HOME directory is:  /nfs/pic.es/user/c/cacosta
My Workdir is:  /home/execute/dir_13420
My PATH is:  /bin:/usr/lib64/qt-3.3/bin:/usr/local/bin:/usr/bin:/usr/local/sbin:/usr/sbin:/nfs/pic.es/user/c/cacosta/bin:/nfs/pic.es/user/c/cacosta/bin
My SOFTWARE directory is:  /software/dteam/

The $ENV(variable) allows access to the environment variables available in the submit server.

• 2) Adding exports in your script

$ cat test2.sh 
#!/bin/bash

export PATH=$PATH:$HOME/bin
export SOFT_DIR=/software/dteam

echo 'My HOME directory is: ' $HOME
echo 'My Workdir is: ' $PWD
echo 'My PATH is: ' $PATH
echo 'My SOFTWARE directory is: ' $SOFT_DIR

$ cat test2.sub 
executable = test2.sh
output = output-$(ClusterId).$(ProcId).out
error = error-$(ClusterId).$(ProcId).err
log = log-$(ClusterId).$(ProcId).log

queue

$ cat output-130.0.out 
My HOME directory is:  /nfs/pic.es/user/c/cacosta
My Workdir is:  /home/execute/dir_15903
My PATH is:  /bin:/usr/local/bin:/usr/bin:/nfs/pic.es/user/c/cacosta/bin
My SOFTWARE directory is:  /software/dteam

Notice that the PATH is different from the first example, it only adds $HOME/bin and not the whole PATH that was loaded with $ENV(PATH).

Queue

Queue is the command in your submit file to select the number of job instances you want to submit. Basically, you can specify the number of jobs of your batch using queue N at the end of your submit file, being N the number of jobs per batch. This is a very powerful tool that allows users to submit several jobs in different ways using the same submission script.

Multiple queue statements

We start with the most simple example. We want to submit 2 jobs using the same executable, each one with their arguments and different cpus.

executable = test.sh

output = test-$(ClusterId).$(ProcId).out
error = test-$(ClusterId).$(ProcId).err
log = test-$(ClusterId).$(ProcId).log

args = -c1 -t 60
request_cpus = 1
queue
args = -c 8 -t 60
request_cpus = 8
queue

Thus, there will be two jobs submitted, the first one using 1 cpu and the second one using 8.

Matching pattern

Another example, we want to submit jobs that match the filenames we have in our current directory.

executable = /bin/echo
arguments = $(filename)
output = test-$(ClusterId).$(ProcId).out
error = test-$(ClusterId).$(ProcId).err
log = test-$(ClusterId).$(ProcId).log

queue filename matching files Hello*

In our current directory:

$ ls Hello*
Hello1  Hello2  Hello3

So, 3 jobs will be submitted.

$ condor_submit test-queue.sub 
Submitting job(s)...
3 job(s) submitted to cluster 321.

$ grep Hello output-321.*out
output-321.0.out:Hello1
output-321.1.out:Hello2
output-321.2.out:Hello3

From file

We have an executable test.sh that can have two arguments -c $(arg1), -y $(arg2) and we want to submit 4 jobs using a different set of arguments. This can be done using queue ... from file.

executable = test.sh
arguments = -c $(arg1) -t $(arg2)
output = test-$(ClusterId).$(ProcId).out
error = test-$(ClusterId).$(ProcId).err
log = test-$(ClusterId).$(ProcId).log

queue arg1,arg2 from arg_list.txt

Where the arg_list.txt is:

$ cat arg_list.txt 
1, 15
2, 10
1, 12
4, 13

In list

Similar that using from file, you can specify your different elements in a list. The next example will submit 4 jobs, each one with the argument specified in the list.

executable = test.sh
arguments = -c 1 -t $(arg1)
output = test-$(ClusterId).$(ProcId).out
error = test-$(ClusterId).$(ProcId).err
log = test-$(ClusterId).$(ProcId).log

queue arg1 in (10 15 5 20)

As you can see there are several ways to use queue command. You can find more examples in the documentation [3].

Transfer files

First of all, note that the standard output and standard error of your job will always be transferred back when the job finishes and here we understand finishing in all the ways: completed jobs, removed jobs and held jobs. However, all the other files generated in the scratch dir of the WN will not be transferred back if you do not specify which files.

transfer_output_files

The next example shows a script called test-output.sh that writes in two files:

#!/bin/bash

for i in {1..10}; do
    echo "Number $i" >> $_CONDOR_SCRATCH_DIR/1.txt
    sleep 2s
    echo "Hello" >> $_CONDOR_SCRATCH_DIR/2.txt
done

This, after submitting this job, the script will generate the file 1.txt and 2.txt in the scratch directory. Note that the condor variable $_CONDOR_SCRATCH_DIR is really useful to be sure that you are pointing to the scratch directory of the node. This is very important if you need to recover any of the files generated in the $_CONDOR_STRACH_DIR of the WN, you should use the option transfer_output_files in your submit file.

executable = test-output.sh
output = test-$(ClusterId).$(ProcId).out
error = test-$(ClusterId).$(ProcId).err
log = test-$(ClusterId).$(ProcId).log
transfer_output_files=1.txt

queue

With transfer_output_files command, you can decide the files you want to be transferred back. Using the same script test-output.sh that generates 1.txt and 2.txt, the 1.txt is the only file which is going to be transferred to the submit machine when the job finishes.

Unfortunately, transfer_output_files does not allow wildcards. Note that if you use a wrong filename, in other words, a file that is not really in the scratch directory, an empty file will be generated after job completion.

Thus, as wildcards are not allowed, the best way to transfer several output files is to use directories. Consider the next script:

$ cat test-output-dir.sh 
#!/bin/bash

cd $_CONDOR_SCRATCH_DIR
mkdir my_files
cd my_files
for i in {1..10}; do
    echo "Number $i" >> ./1.txt
    sleep 2s
    echo "Hello" >> ./2.txt
done

We are creating a local directory in our scratch dir called my_files and putting there the 1.txt and 2.txt files. Then, we transfer back the whole directory because we are using transfer_output_files=my_files:

executable = test-output-dir.sh
output = test-out1-$(ClusterId).$(ProcId).out
error = test-out1-$(ClusterId).$(ProcId).err
log = test-out1-$(ClusterId).$(ProcId).log
transfer_output_files=my_files
queue

In our submit directory, after the job finishes, we are going to find the directory with the files.

$ ls my_files/
1.txt  2.txt

Of course, if you only want one file from the created directory, you just need to use transfer_output_files=my_files/2.txt for instance. In that case, only the file will be transferred, the directory is not going to be created in the submit side.

transfer_output_remaps

On the other hand, if you want that to transfer the output file to another directory, you should use transfer_output_remaps option.

executable = test-output.sh
output = test-$(ClusterId).$(ProcId).out
error = test-$(ClusterId).$(ProcId).err
log = test-$(ClusterId).$(ProcId).log
transfer_output_files=1.txt
transfer_output_remaps="1.txt=outputs/1-$(ClusterId).$(ProcId).txt"

queue

The submit file above is going to transfer the 1.txt file from the scratch dir to ./outputs/1-$(ClusterId).$(ProcId).txt

transfer_executable

The Transfer executable mechanism defaults to False for all HTCondor jobs. This means that the executable is searched in the job destination, in other words, the WN. If for some reason, this configuration does not fit your needs, please contact our support team.

Accounting Group

The priority of your job is calculated depending on the Accounting Group you belong to. The common users do not have to worry about the Accounting Group, as it will be automatically taken considering the primary group.

Anyway, if you are in two groups and need to change your Accounting Group for any submission, you can add +experiment="experiment" option in your submit file. Thus, for instance, there is one user that has main group vip and secondary group virgo and want to submit a job that only accounts to virgo.

executable = test.sh
args = --cpu 1 --timeout 120
output = test-$(ClusterId).$(ProcId).out
error = test-$(ClusterId).$(ProcId).err
log = test-$(ClusterId).$(ProcId).log

+experiment="virgo"

queue

Using the option +experiment="virgo" the job will have the share of the virgo experiment and will also be accounted in our records as a virgo job.

The local experiments available now at PIC are:

• atlas
• co2flux
• cta
• des
• desi
• euclid
• magic
• mice
• neutrinos
• paus
• vip
• virgo

Furthermore, the automatic assignation of the Accounting Groups is also done for those groups that have dedicated User Interfaces (such as mic.magic, at3 and ui01-virgo). In that case, the Accounting Group is created according to the User Interface group owner. In other words, it does not matter the main group of the user, if I submit a job from mic.magic machine, my AccountingGroup belongs to magic.

Interactive submission

As it happens with Torque/Maui, HTCondor also has the possibility to submit interactive jobs.

$ condor_submit -interactive

You have to wait in queue and finally you enter inside a WN:

$ condor_submit -interactive
Submitting job(s).
1 job(s) submitted to cluster 52166.
Waiting for job to start...
Welcome to slot1_9@td730.pic.es!
You will be logged out after 7200 seconds of inactivity.
[cacosta@td730 dir_23132]$ 

Notice that you dispose of 7200 seconds of inactivity. Anyway, you can use a submit file to launch an interactive job:

$ condor_submit -interactive test.sub

The session created in the node is affected by the same restrictions of cpu, memory, disk, etc. However, there are options of the submit file that has no sense in an interactive session: executable, arguments or queue.

Notifications

You can enable e-mail notifications in your HTCondor jobs. This is not enabled as default and notifications have to be considered carefully because if you submit several jobs you can receive too many emails. You need to specify the notification and the notify_user attributes in your submit file:

executable = test.sh
arguments = -c 1 -t 60 --vm 4 --vm-bytes 10G
output = test-$(ClusterId).$(ProcId).out
error = test-$(ClusterId).$(ProcId).err
log = test-$(ClusterId).$(ProcId).log

notification = Always
notify_user = xxxxx@xxx.xx

queue

There are 3 options for notification attribute:

• Always. Notifications will be sent when the job is held or completed. This is the required option if you want to know when the job is put on hold state due to the limits of the flavours. Take into account that if the user is the one who removes the job, no notification will be sent.
• Complete. You will receive the notification only for terminated jobs, so, if the job is put on hold by a PeriodicHold expression, no notification will be sent.
• Error. A notification will be sent for jobs that end abnormally, this means that your script is given an Exit Status different from 0 or if the job is placed on hold because of failure (for instance, the log cannot be written in the desired location).

The notify_user attribute is used to tell HTCondor your e-mail address.

Dagman

HTCondor Dagman (Directed Acyclic Graph Manager) is a meta-scheduler for HTCondor [4]. In other words, you can submit to the queue a scheduler that manages the execution order, the hierarchy, of several jobs. This is the best tool to handle multiple jobs interdependent.

For instance, we have two jobs, A and B, where B depends directly on A, this means that we have to be sure that A is finished before submitting B. This is the simplest use case of a dagman workflow.

Next, you can see the 2 jobs. Job A consists of writing several files in your scratch dir and job B to remove it. Note that we need to create one submit file per each job.

$ cat jobA.sh
#!/bin/bash

ARG1=$1

cd $_CONDOR_SCRATCH_DIR
mkdir test
cd test
cat << EOF > ./dagmanfile${ARG1}.test
EOF

sleep 30

$ cat jobB.sh 
#!/bin/bash

for a in $(seq 1 100); do
    rm -f $HOME/condor/dagman/test/dagmanfile${a}.test
    echo "dagmanfile$a.test erased" 
done

sleep 30

And here you have the 2 submit files. Take into account that we are transferring back the test directory from scratch to our submit path. The job B is going to remove these files later.

$ cat dagjobA.sub 
executable = jobA.sh
args = $(Item)
output = jobA-output-$(ClusterId).$(ProcId).out
error = jobA-error-$(ClusterId).$(ProcId).err
log = jobA-log-$(ClusterId).$(ProcId).log
transfer_output_files=test
queue from seq 1 1 100 |

$ cat dagjobB.sub 
executable = jobB.sh
output = jobB-output-$(ClusterId).$(ProcId).out
error = jobB-error-$(ClusterId).$(ProcId).err
log = jobB-log-$(ClusterId).$(ProcId).log
queue

Then, there is the dagman file:

$ cat dagjob.dag 
JOB  A  dagjobA.sub
JOB  B  dagjobB.sub
PARENT A CHILD B

To submit the dagman file you should use condor_submit_dag.

$ condor_submit_dag dagjob.dag 

-----------------------------------------------------------------------
File for submitting this DAG to HTCondor           : dagjob.dag.condor.sub
Log of DAGMan debugging messages                 : dagjob.dag.dagman.out
Log of HTCondor library output                     : dagjob.dag.lib.out
Log of HTCondor library error messages             : dagjob.dag.lib.err
Log of the life of condor_dagman itself          : dagjob.dag.dagman.log

Submitting job(s).
1 job(s) submitted to cluster 60771.
-----------------------------------------------------------------------

Several files are generated: the submit file to HTCondor of your dagman jobs, the log, the library standard output and standard error messages and the log file of the condor_dagman. Take into account that this standard output and error consist in the dagman execution, the regular standard output and error of your jobs will be created as normal HTCondor jobs.

Querying with the condor_q we can try to understand what is happening.

$ condor_q 60771


-- Schedd: submit01.pic.es : <193.109.174.82:9618?... @ 07/05/19 16:21:39
OWNER   BATCH_NAME          SUBMITTED   DONE   RUN    IDLE  TOTAL JOB_IDS
cacosta dagjob.dag+60771   7/5  16:21      _    100      _      1 60772.0-99

Total for query: 100 jobs; 0 completed, 0 removed, 0 idle, 100 running, 0 held, 0 suspended 
Total for all users: 920 jobs; 148 completed, 0 removed, 632 idle, 140 running, 0 held, 0 suspended

The dagman scheduler has Id 60771, then the 100 first jobs generated by the jobA have the ClusterId.ProcId 60772.0-99.

$ condor_q 60771


-- Schedd: submit01.pic.es : <193.109.174.82:9618?... @ 07/05/19 16:22:53
OWNER   BATCH_NAME          SUBMITTED   DONE   RUN    IDLE  TOTAL JOB_IDS
cacosta dagjob.dag+60771   7/5  16:21      _      1      _      1 60773.0

Total for query: 1 jobs; 0 completed, 0 removed, 0 idle, 1 running, 0 held, 0 suspended 
Total for all users: 816 jobs; 144 completed, 0 removed, 631 idle, 41 running, 0 held, 0 suspended

After the 100 instances of job A are finished, you can see a test directory in your submit path with several dagmanfilesX.file. Then, the job B starts.

jobA and jobB work as regular HTCondor jobs (flavours, requests, etc.). The standard output, error and log files for all the jobs will be generated. Furthermore, there is the different dagman files we have commented above.

Think in a more complex workflow like the next one:

The dagman file has to be something like this:

JOB  A  jobA.sub
JOB  B  jobB.sub
JOB  C  jobC.sub
JOB  D  jobD.sub
JOB  E  jobE.sub
JOB  F  jobF.sub
JOB  G  jobG.sub
JOB  H  jobH.sub
PARENT A CHILD B C D
PARENT B C CHILD E
PARENT D CHILD F G H

SCRIPT

There is the SCRIPT command used to specify any processing work you need to execute pre or post the jobs finish. For instance, PRE is used to place files somewhere and POST to clean up files. Take into account that if your scripts give an exit value different from 0 the workflow is going to stop.

A simple example. jobA creates several files in the SCRATCH and jobB add a number in each file.

$ cat jobA.sh 
#!/bin/bash

ARG1=$1
cd $_CONDOR_SCRATCH_DIR
mkdir test
cd test/
touch ./dagmanfile${ARG1}.test

sleep 30

$ cat jobB.sh 
#!/bin/bash

cd $_CONDOR_SCRATCH_DIR
for a in $(seq 1 100); do
    echo $a > $HOME/condor/dagman/test/dagmanfile${a}.test
done

sleep 30

$ cat dagjobA.sub 
executable = jobA.sh
args = $(Item)
output = jobA-output-$(ClusterId).$(ProcId).out
error = jobA-error-$(ClusterId).$(ProcId).err
log = jobA-log-$(ClusterId).$(ProcId).log
transfer_output_files=test
queue from seq 1 1 100 |

$ cat dagjobB.sub
executable = jobB.sh
output = jobB-output-$(ClusterId).$(ProcId).out
error = jobB-error-$(ClusterId).$(ProcId).err
log = jobB-log-$(ClusterId).$(ProcId).log
queue 

After job B execution, we want to create a tar file and remove the created test/ directory.

$ cat tar.sh
#!/bin/bash

tar -cvf file-test.tar test/*.test
rm -rf test

Thus, this is the dagman file we need:

$ cat dagjob-SCRIPT.dag 
JOB  A  dagjobA.sub
JOB  B  dagjobB.sub
SCRIPT POST B tar.sh
PARENT A CHILD B

We submit the job:

$ condor_submit_dag dagjob-SCRIPT.dag 

-----------------------------------------------------------------------
File for submitting this DAG to HTCondor           : dagjob-SCRIPT.dag.condor.sub
Log of DAGMan debugging messages                 : dagjob-SCRIPT.dag.dagman.out
Log of HTCondor library output                     : dagjob-SCRIPT.dag.lib.out
Log of HTCondor library error messages             : dagjob-SCRIPT.dag.lib.err
Log of the life of condor_dagman itself          : dagjob-SCRIPT.dag.dagman.log

Submitting job(s).
1 job(s) submitted to cluster 60807.
-----------------------------------------------------------------------

Monitor the behavior.

$ condor_q


-- Schedd: submit01.pic.es : <193.109.174.82:9618?... @ 07/08/19 10:28:15
OWNER   BATCH_NAME                 SUBMITTED   DONE   RUN    IDLE  TOTAL JOB_IDS
cacosta dagjob-SCRIPT.dag+60807   7/8  10:28      _      _      _      1 0.0

Total for query: 0 jobs; 0 completed, 0 removed, 0 idle, 0 running, 0 held, 0 suspended 
Total for cacosta: 0 jobs; 0 completed, 0 removed, 0 idle, 0 running, 0 held, 0 suspended 
Total for all users: 223 jobs; 0 completed, 0 removed, 0 idle, 223 running, 0 held, 0 suspended

$ condor_q


-- Schedd: submit01.pic.es : <193.109.174.82:9618?... @ 07/08/19 10:28:18
OWNER   BATCH_NAME                 SUBMITTED   DONE   RUN    IDLE  TOTAL JOB_IDS
cacosta dagjob-SCRIPT.dag+60807   7/8  10:28      _      _    100      1 60808.0-99

Total for query: 100 jobs; 0 completed, 0 removed, 100 idle, 0 running, 0 held, 0 suspended 
Total for cacosta: 100 jobs; 0 completed, 0 removed, 100 idle, 0 running, 0 held, 0 suspended 
Total for all users: 323 jobs; 0 completed, 0 removed, 100 idle, 223 running, 0 held, 0 suspended

After the job B finishes, a files-test.tar file is created and the test directory removed.

Monitoring your jobs

The basic tools to monitor your jobs by command line interface are condor_q (the principal one), condor_wait, condor_history, condor_tail and condor_ssh_to_job.

condor_q

The condor_q command is the one to query the queue of the schedd [5]. As other commands in HTcondor, condor_q allows to specify clearly what you want using the "-constraint" option to filter the Job Attributes you want to query. To know all the jobs attributes, you use condor_q -l job_id. Furthermore, you can tune the output with "-autoformat" or "-af" option.

It is better to show the potential of condor_q in an example:

$ condor_q -const "RequestCpus > 1 && JobStatus == 1" -nobatch


-- Schedd: submit01.pic.es : <193.109.174.82:9618?... @ 03/14/19 09:42:46
 ID      OWNER            SUBMITTED     RUN_TIME ST PRI SIZE CMD
 630.0   cacosta         3/14 09:42   0+00:00:00 I  0    0.0 test.sh --cpu 1 --timeout 10s
 630.1   cacosta         3/14 09:42   0+00:00:00 I  0    0.0 test.sh --cpu 1 --timeout 10s
 630.2   cacosta         3/14 09:42   0+00:00:00 I  0    0.0 test.sh --cpu 1 --timeout 10s
 630.3   cacosta         3/14 09:42   0+00:00:00 I  0    0.0 test.sh --cpu 1 --timeout 10s
 630.4   cacosta         3/14 09:42   0+00:00:00 I  0    0.0 test.sh --cpu 1 --timeout 10s

Total for query: 5 jobs; 0 completed, 0 removed, 5 idle, 0 running, 0 held, 0 suspended
Total for cacosta: 5 jobs; 0 completed, 0 removed, 5 idle, 0 running, 0 held, 0 suspended 
Total for all users: 7 jobs; 0 completed, 0 removed, 5 idle, 2 running, 0 held, 0 suspended

We use the constraint (-const) to filter our jobs. Here, the query filter for the jobs that request more than one cpu, that are Idle (JobStatus ==1) and using -nobatch the command shows your jobs ungrouped.

On the other hand, filtering the same jobs, we can decide the format of the output just showing the job attributes we are interested in.

$ condor_q -const "RequestCpus > 1 && JobStatus == 1" -nobatch -af ClusterId ProcId RequestCpus RequestMemory
630 0 4 2048
630 1 4 2048
630 2 4 2048
630 3 4 2048
630 4 4 2048

analyze and better-analyze

Furthermore, the condor_q command allows the options -analyze and -better-analyze that show you the reason why your job is not running.

$ condor_q -analyze 146.0


-- Schedd: condor-ui01.pic.es : <193.109.175.231:9618?...
The Requirements expression for job 146.000 is

    (TARGET.WN_property == ifThenElse(MY.WN_property is undefined,"default",MY.WN_property)) && (TARGET.Arch == "X86_64") && (TARGET.OpSys == "LINUX") &&
    (TARGET.Disk >= RequestDisk) && (TARGET.Memory >= RequestMemory) && ((TARGET.FileSystemDomain == MY.FileSystemDomain) || (TARGET.HasFileTransfer))



146.000:  Job has not yet been considered by the matchmaker.


146.000:  Run analysis summary ignoring user priority.  Of 276 machines,
      3 are rejected by your job's requirements
      1 reject your job because of their own requirements
      0 match and are already running your jobs
      0 match but are serving other users
    272 are able to run your job

$ condor_q -better-analyze 148


-- Schedd: condor-ui01.pic.es : <193.109.175.231:9618?...
The Requirements expression for job 148.000 is

    (TARGET.WN_property == ifThenElse(MY.WN_property is undefined,"default",MY.WN_property)) && (TARGET.Arch == "X86_64") && (TARGET.OpSys == "LINUX") &&
    (TARGET.Disk >= RequestDisk) && (TARGET.Memory >= RequestMemory) && ((TARGET.FileSystemDomain == MY.FileSystemDomain) || (TARGET.HasFileTransfer))

Job 148.000 defines the following attributes:

    FileSystemDomain = "condor-ui01.pic.es"
    RequestCpus = 1
    RequestDisk = 20971520 * RequestCpus
    RequestMemory = 2048 * RequestCpus

The Requirements expression for job 148.000 reduces to these conditions:

         Slots
Step    Matched  Condition
-----  --------  ---------
[0]        3761  TARGET.WN_property == ifThenElse(MY.WN_property is undefined,"default",MY.WN_property)
[5]         277  TARGET.Disk >= RequestDisk
[6]         274  [0] && [5]
[7]        3744  TARGET.Memory >= RequestMemory
[8]         253  [6] && [7]
[10]       3765  TARGET.HasFileTransfer


148.000:  Job has not yet been considered by the matchmaker.


148.000:  Run analysis summary ignoring user priority.  Of 276 machines,
      3 are rejected by your job's requirements
      1 reject your job because of their own requirements
      0 match and are already running your jobs
      0 match but are serving other users
    272 are able to run your job

The jobs in these examples are not considered by the matchmaker yet but you can see that there are 272 machines available that can run your job.

Useful Job Attributes

There are several Job Attributes in your job. Here you have a list of few of them:

• JobStatus. Number indicating the status of your job. Relevant Job Status numbers:

JobStatus	Name	Symbol	Description
1	Idle	I	Job is idle, queued waiting for resources
2	Running	R	Job is running
3	Removed	X	Job has been removed by user or admin
4	Completed	C	Job is completed
5	Held	H	Job is in hold state, it will not be scheduled until released

• RemoteHost. The WN where the jobs are running.
• ResidentSetSize_RAW. The maximum observed physical memory consumed by the job in KiB while running.
• DiskUsage_RAW. The maximum observed disk usage by the job in KiB while running.
• RemoteUserCpu. Total number of seconds of user CPU time the job has used.
• Owner. The submitter user.
• OwnerGroup. Main group or experiment of the submitter user.

There are other HTCondor "magic" numbers that you can consult [6].

condor_wait

The condor_wait command allows us to watch and extract information from the user log file [7]. This command waits forever until the job is finished unless a wait time is specified (with -wait option). Furthermore, as condor_wait monitors the log file, it requires a job successfully submitted to be executed.

It is not as useful as condor_q but can give you information about several jobs if you collect them in the same log file.

For instance, monitoring 3 jobs:

$ condor_wait -status log-test.log 
176.0.0 submitted
176.1.0 submitted
176.2.0 submitted
176.1.0 executing on host <192.168.100.15:9618?addrs=192.168.100.15-9618+[2001-67c-1148-301--213]-9618&noUDP&sock=24729_9d98_3>
176.2.0 executing on host <192.168.100.173:9618?addrs=192.168.100.173-9618+[2001-67c-1148-301--73]-9618&noUDP&sock=14012_a51f_3>
176.0.0 executing on host <192.168.101.68:9618?addrs=192.168.101.68-9618&noUDP&sock=2583_4d37_3>
176.1.0 completed
176.2.0 completed
176.0.0 completed
All jobs done.

condor_history

Once the job is removed from the queue, you can query it using condor_history [8]. The condor_history command allows similar constraints than condor_q. The jobs kept in history are from today until to the previous month. If you need information for an older job, please ask your contact.

It is recommended to use the option "-limit N" where N is the number of jobs you want to query to perform faster queries.

$ condor_history -const 'Owner == "cacosta" && JobStatus ==4' -limit 5
 ID     OWNER          SUBMITTED   RUN_TIME     ST COMPLETED   CMD            
  78.0   cacosta         3/14 10:05   0+00:01:42 C   3/14 10:07 /nfs/pic.es/user/c/cacosta/condor/test-local/remote/test.sh --cpu 1 --timeout 100s
  76.0   cacosta         3/14 10:03   0+00:01:41 C   3/14 10:04 /nfs/pic.es/user/c/cacosta/condor/test-local/remote/test.sh --cpu 1 --timeout 100s
  75.0   cacosta         3/14 09:59   0+00:01:45 C   3/14 10:01 /nfs/pic.es/user/c/cacosta/condor/test-local/remote/test.sh --cpu 1 --timeout 100s
  74.0   cacosta         3/14 09:13   0+00:00:25 C   3/14 09:14 /nfs/pic.es/user/c/cacosta/condor/test-local/remote/test-output.sh 
  73.0   cacosta         3/14 09:07   0+00:00:22 C   3/14 09:08 /nfs/pic.es/user/c/cacosta/condor/test-local/remote/test-output.sh 

condor_tail

The condor_tail command checks the standard output and error of job while this is running [9].

$ condor_tail -f 65.0
stress: info: [9] dispatching hogs: 1 cpu, 0 io, 0 vm, 0 hdd

The default is to check the standard output and the "-f" option, follow, acts as tail -f Linux command, repetitively tail the file until interrupted. To check the standard error you need to use "-stderr" option.

condor_ssh_to_job

Finally, another way to monitor how your job is evolving is the condor_ssh_to_job command [10]. Using this command, the user can enter into the job directory of the node and check what is happening.

$ condor_ssh_to_job 66.0
Welcome to slot1_7@td622.pic.es!
Your condor job is running with pid(s) 577 1163.
$ ls
condor_exec.exe  _condor_stderr  _condor_stdout  out.txt  tmp  var

Removing your jobs

To remove your jobs, you have to use condor_rm command [11].

The most common way to use condor_rm is just specifying the ClusterId and/or the ProcId of your job. For instance, we have a batch of 4 jobs.

$ condor_q 154 -nobatch


-- Schedd: condor-ui01.pic.es : <193.109.175.231:9618?... @ 04/01/19 15:37:55
 ID      OWNER            SUBMITTED     RUN_TIME ST PRI SIZE CMD
 154.0   cacosta         4/1  15:37   0+00:00:00 I  0    0.0 test.sh --timeout 600s --vm 2 --vm-bytes 2G
 154.1   cacosta         4/1  15:37   0+00:00:00 I  0    0.0 test.sh --timeout 600s --vm 2 --vm-bytes 2G
 154.2   cacosta         4/1  15:37   0+00:00:00 I  0    0.0 test.sh --timeout 600s --vm 2 --vm-bytes 2G
 154.3   cacosta         4/1  15:37   0+00:00:00 I  0    0.0 test.sh --timeout 600s --vm 2 --vm-bytes 2G

Total for query: 4 jobs; 0 completed, 0 removed, 4 idle, 0 running, 0 held, 0 suspended 
Total for all users: 7 jobs; 0 completed, 2 removed, 4 idle, 0 running, 1 held, 0 suspended

Use ClusterId.ProcId to remove just one job of the cluster:

$ condor_rm 154.2
Job 154.2 marked for removal

Use ClusterId to remove all the jobs in a cluster:

$ condor_rm 154
All jobs in cluster 154 have been marked for removal

You can use constraints to remove all the jobs that meet any condition.

$ condor_rm -const 'RequestCpus > 1'
All jobs matching constraint (RequestCpus > 1) have been marked for removal

Or you can remove all your jobs just using the "-all" option.

From Torque to HTCondor

Although you can find more commands in HTCondor than the ones you are used in Torque/PBS, the most common commands in Torque (qsub, qdel and qstat) have their equivalent in HTCondor.

Torque	HTCondor	Description
qsub	condor_submit	To submit jobs to the farm
qdel	condor_rm	To remove your job or all the jobs from an user
qstat	condor_q	To query the state of your jobs

HTCondor has a powerful language to query the pool and more and interesting options to monitor your job that has not their equivalent in Torque/Maui (condor_history for instance), thus, do not hesitate to look into the HTCondor documentation to create your queries and learn more about the commands.

Remember that there is no concept of differentiated queues with static requirements in HTCondor. Therefore, executing condor_q you are querying your jobs in the whole schedd, do not expect the different queues showed as with qstat command. At PIC we use the flavour option to limit the walltime of the jobs in a similar way as they were limited by the queues in Torque/Maui.

Issues

Why are my jobs held?

There are several reasons why your jobs can be held. You can ask HTCondor why your jobs are held using the next command:

$ condor_q -const 'JobStatus == 5' -af HoldReason

The job attribute PeriodicHold expression ... evaluated to TRUE

You will find several times this reason:

The job attribute PeriodicHold expression '(JobStatus == 2 && (time() - EnteredCurrentStatus) > 172800) || (JobStatus == 2 && ifThenElse(ResidentSetSize =!= undefined,ResidentSetSize > JobMemoryLimit * 1024,false))' evaluated to TRUE

This is a Hold Reason controlled by the system administrator. On one hand, there is the time your job has been running (10800, 172800 or 345600 seconds depending in your flavour, short, medium or long respectively). The second part of the PeriodicHold expression refers to the memory used by your job. If your job consumes 1.5 times the memory requested, it will be put on hold.

Thus, if your job is on hold and the elapsed time was not enough, you can query the memory in the next way:

# condor_q 265321.8 -af HoldReason
The job attribute PeriodicHold expression '(JobStatus == 2 && (time() - EnteredCurrentStatus) > 172800) || (JobStatus == 2 && ifThenElse(ResidentSetSize =!= undefined,ResidentSetSize > JobMemoryLimit * 1024,false))' evaluated to TRUE

# condor_q 265321.8 -af ResidentSetSize JobMemoryLimit*1024
3250000 3145728.0

Clearly, your job is consuming more memory than requested (exceeding 1.5 the memory requested!). You should resubmit your job increasing the memory requested.

CPU usage exceeded request_cpus

The HoldReason, in this case, means that your job is taking more cpus than requesting for the last hour. You should resubmit your job increasing the number of requested cpus.

Why are my jobs idle?

This is one of the most typical questions when you submit jobs to a batch system. You have always to assume that your job will be idle for some time and this time is not always the same, it depends in the requirements of your jobs, the share of your group and the total use of the farm.

Typically, if your needs are smaller (1 CPU and 1 GB of RAM for instance) your job will be idle for less time than if you need 8 CPUs or more than 2 GB RAM per core.

Anyway, remember that you have the options -analyze and -better-analyze in condor_q command that gives you information about why your job is idle (without taking into account the priority). These options are especially useful to detect jobs that are Idle because the requirements are not met.

https://pwiki.pic.es/index.php?title=HTCondor_User_Guide#analyze_and_better-analyze

When your job is requirements cannot match the WorkerNodes available, you will see a WARNING message like this using -analyze or better-analyze

$ condor_q -analyze 267201


-- Schedd: submit01.pic.es : <193.109.174.82:9618?...
The Requirements expression for job 267201.000 is

    (TARGET.WN_property == ifThenElse(MY.WN_property is undefined,"default",MY.WN_property)) && (TARGET.Arch == "X86_64") && (TARGET.OpSys == "LINUX") &&
    (TARGET.Disk >= RequestDisk) && (TARGET.Memory >= RequestMemory) && (TARGET.Cpus >= RequestCpus) && ((TARGET.FileSystemDomain == MY.FileSystemDomain) ||
      (TARGET.HasFileTransfer))


No successful match recorded.
Last failed match: Fri Feb  7 16:58:04 2020

Reason for last match failure: no match found 

267201.000:  Run analysis summary ignoring user priority.  Of 338 machines,
    338 are rejected by your job's requirements
      0 reject your job because of their own requirements
      0 match and are already running your jobs
      0 match but are serving other users
      0 are able to run your job

WARNING:  Be advised:
   No machines matched the jobs's constraints

References and links

You can find many documentation about HTCondor. Here you have a list of useful links from this manual.

• [1] https://htcondor.readthedocs.io/en/v8_8_3/users-manual/index.html

• [2] condor_submit: https://htcondor.readthedocs.io/en/v8_8_3/users-manual/submitting-a-job.html

• [3] Queue command: https://htcondor.readthedocs.io/en/v8_8_3/users-manual/submitting-a-job.html

• [4] https://htcondor.readthedocs.io/en/v8_8_3/man-pages/condor_q

• [5] HTCondor magic numbers: https://htcondor-wiki.cs.wisc.edu/index.cgi/wiki?p=MagicNumbers

• [6] condor_wait: https://htcondor.readthedocs.io/en/v8_8_3/man-pages/condor_wait.html

• [7] condor_history: https://htcondor.readthedocs.io/en/v8_8_3/man-pages/condor_history.html

• [8] condor_tail: https://htcondor.readthedocs.io/en/v8_8_3/man-pages/condor_tail.html

• [9] condor_ssh_to_job: https://htcondor.readthedocs.io/en/v8_8_3/man-pages/condor_ssh_to_job.html

• [10] condor_rm https://htcondor.readthedocs.io/en/v8_8_3/man-pages/condor_rm.html

• HTcondor User Guide Tutorial: https://docs.google.com/presentation/d/1-64fEcfLyxLzSpZH-tbV-SjMAc0CvMpWhE0I4oaDKkQ/edit?usp=sharing

• HTCondor Tutorial examples: https://github.com/PortdInformacioCientifica/htcondor-tutorial