The DAQInterface Manual¶
- Table of contents
- The DAQInterface Manual
- A brief overview
- Getting started
- A first walkthrough
- A second walkthrough
- DAQInterface for your experiment
- Examining your output
This manual was last edited on Jul-29-2019 and covers DAQInterface v3_06_00. For instructions on different versions of DAQInterface, click on one of the links below:
v3_04_01 and v3_05_00
v3_03_00 and v3_03_01
develop (not always guaranteed to be up-to-date)
Instructions for developers and for special debugging scenarios are here
A brief overview¶
DAQInterface is a set of Python and Bash scripts which make it easy to start and run an existing artdaq-based DAQ system. Specifically, this means the ability to select how many of each type of artdaq process (BoardReaders, EventBuilders, DataLoggers, Dispatchers and RoutingMasters) to use in a running DAQ system, as well as what hosts they should run on. Additionally, DAQInterface provides built-in support for sending transitions via XML-RPC to artdaq processes at the command line as well as the saving of metadata related to a run (such as the version of artdaq used, the location of the artdaq logfiles, etc.).
Getting DAQInterface to perform runs takes about 10 minutes, assuming you already have an installation of the artdaq-demo package which is based on artdaq v3. If you don't already have an artdaq-demo installation, you can just execute the following:
wget https://cdcvs.fnal.gov/redmine/projects/artdaq-demo/repository/revisions/develop/raw/tools/quick-mrb-start.sh chmod +x quick-mrb-start.sh ./quick-mrb-start.sh --tag=v3_06_00
For using other artdaq-based packages besides artdaq-demo with DAQInterface, after reading this wiki take a look at the links at the bottom of this page. If you're stuck after reading the documentation, please contact John Freeman at email@example.com.
If you're using a ups products directory which doesn't yet contain DAQInterface v3_06_00, you can get it from Fermilab's SciSoft site:
cd <products directory> # Replace <products directory> with the actual name of your products directory curl -O https://scisoft.fnal.gov/scisoft/packages/artdaq_daqinterface/v3_06_00/artdaq_daqinterface-3.06.00-noarch.tar.bz2 tar xjf artdaq_daqinterface-3.06.00-noarch.tar.bz2 rm -f artdaq_daqinterface-3.06.00-noarch.tar.bz2
And when you have DAQInterface v3_06_00 in your products directory, you can set it up:
source <products directory>/setup # Replace <products directory> with the actual name of your products directory setup artdaq_daqinterface v3_06_00
Now, let's create a DAQInterface work area:
mkdir $HOME/DAQInterface # You can, of course, make a directory with another name if you like cd $HOME/DAQInterface cp $ARTDAQ_DAQINTERFACE_DIR/bin/mock_ups_setup.sh . cp $ARTDAQ_DAQINTERFACE_DIR/docs/user_sourcefile_example . cp $ARTDAQ_DAQINTERFACE_DIR/docs/settings_example . cp $ARTDAQ_DAQINTERFACE_DIR/docs/known_boardreaders_list_example . cp $ARTDAQ_DAQINTERFACE_DIR/docs/boot.txt .
Since DAQInterface doesn't know the experiment-specific requirements of its user, we'll want to create a file to source which sets up an environment appropriate for our needs. In our area, this file is "user_sourcefile_example". Let's open it up, and look at the top line:
What you'll want to do is change this top line so that the DAQINTERFACE_SETTINGS variable refers to fully-qualified filename of the settings_example file we copied into our area. Once you've done that, let's open the settings_example file itself, and modify the line
so that the value of "productsdir_for_bash_scripts" points to a products directory which contains xmlrpc_c (for sending messages to DAQInterface),root (to allow the execution of scripts which crosscheck the *.root files output by the DAQ), and fhiclcpp (so DAQInterface can reformat the FHiCL documents it sends to the artdaq processes). Then modify the lines
log_directory: /tmp record_directory: $HOME/run_records
so they refer to the directories in which you'd like artdaq to output its logfiles (
log_directory) and to save records of what happened during each run (
record_directory). It's recommended that you not leave the log directory as "/tmp", otherwise your logfiles will most likely be periodically removed. Of course, make sure the directories you set here actually exist.
The last file we'll edit is the boot.txt file. This file will be passed to DAQInterface on the boot transition (i.e., the transition in which DAQInterface launches the artdaq processes). Take its top line:
DAQ setup script: name_of_DAQ_setup_script_not_yet_defined
...and replace name_of_DAQ_setup_script_not_yet_defined with the fully-qualified filename of the script you want DAQInterface to source before it tries to launch artdaq processes. E.g., if you installed artdaq-demo with the quick-mrb-start.sh script in $HOME/artdaq-demo-installation, you'd want to replace "name_of_DAQ_setup_script_not_yet_defined" with "$HOME/artdaq-demo-installation/setupARTDAQDEMO".
You'll also see that hosts (all
localhost) have been defined for two EventBuilders, a DataLogger and a Dispatcher; we'll leave them like that for now.
A first walkthrough¶
Now that we've made the needed modifications to the files in our area, it's time to set up the environment for DAQInterface to run. In the terminal, type
source <products directory>/setup # If you didn't do this earlier setup artdaq_daqinterface v3_06_00 # " " " " " " export DAQINTERFACE_USER_SOURCEFILE=$PWD/user_sourcefile_example source $ARTDAQ_DAQINTERFACE_DIR/source_me
...where for the future, you may want to consider putting all those commands in a single file to source, since you need to do this whenever you log in and want to use DAQInterface. You'll see something like the following:
* The command was successful * To launch, just type "DAQInterface &" (excluding quotes, ampersand optional) * Output will be logged in /tmp/daqinterface_jcfree/DAQInterface_partition<N>.log, where <N> is the partition the DAQInterface instance is on * Help is available at https://cdcvs.fnal.gov/redmine/projects/artdaq-utilities/wiki/Artdaq-daqinterface
so then we type
and you'll see one of two possible outputs. Assuming there isn't already a DAQInterface running on the default port DAQInterface listens on, you'll see something like the following:
Mon Jul 29 12:18:54 CDT 2019: DAQInterface in partition 0 launched and now in "stopped" state, listening on port 10000
in which case the launch was successful, and you can skip down to the next paragraph. However, if there IS a DAQInterface already running on your host, you'll see:
There already appears to be a DAQInterface instance running on the requested partition number (0); please either kill the instance (if it's yours) or use a different partition. Run "listdaqinterfaces.sh" for more info.
where if you run
listdaqinterfaces.shyou'll see something like:
DAQInterface instance was launched at 11:08 by jcfree in partition 0 listening on port 10000
In your environment, there's an environment variable called DAQINTERFACE_PARTITION_NUMBER which is used when DAQInterface is launched to determine which port to listen on. Partition numbers count up sequentially from 0, and only one DAQInterface instance can be used per partition. The default value of DAQINTERFACE_PARTITION_NUMBER is 0; if there's already a DAQInterface instance running on this partition, your options are (1) to launch DAQInterface on an unused partition, or (2) to kill the DAQInterface instance using partition 0, assuming it's controlled by your user account and you know it's not in active use. To launch DAQInterface in an unused partition, you could just do the following, e.g.:
export DAQINTERFACE_PARTITION_NUMBER=1 DAQInterface &
and to kill the existing DAQInterface on partition 0 and launch your new one, you could do
kill_daqinterface_on_partition.sh 0 DAQInterface &
...where if the kill_daqinterface_on_partition.sh command appears to hang, hit Ctrl-c and run it again with the option "--force" added. DAQInterface is now ready to receive transition commands. Now type the following:
just_do_it.sh ./boot.txt 0 --config "demo" --comps "component01 component02"
This command can be understood to mean, "Give DAQInterface the file ./boot.txt which tells it how many artdaq processes to launch and where to launch them, initialize the processes after they're launched with the FHiCL documents found in the configuration called demo, and in particular, initialize two BoardReaders with FHiCL documents called component01.fcl and component02.fcl". After executing the command, give it a few moments, and you should eventually see something like the following:
Mon Sep 18 14:29:13 CDT 2017: START transition complete for run 959 Will acquire data until Ctrl-C is hit
DAQInterface is now in the "running" state. We'll step through the commands executed automatically by the script in a moment, but essentially what we've done is launched a system running two ToySimulator fragment generators in two BoardReader processes, where events are assembled by two EventBuilders and sent to a DataLogger which saves them to disk and sends them to a Dispatcher. To stop datataking, hit
Ctrl-cfrom the terminal running
just_do_it.sh. Momentarily, you should return to the command line, and in the terminal running DAQInterface, you may need to scroll up a bit but you should see something like:
Wed Dec 12 17:24:50 CST 2018: TERMINATE transition complete Process manager logfiles (if applicable): mu2edaq01.fnal.gov:/home/jcfree/daqlogs/pmt/pmt-1740.1-20181212172243.log
A second walkthrough¶
just_do_it.sh strung multiple transitions together in order to make life simpler for the user, for educational purposes it's useful to step through running DAQInterface one transition at a time. In this section, we'll do just that, as well as provide a fuller explanation for the output produced during a run. While not required, you may want to execute the commands from another terminal as it makes the interface a bit cleaner; in order to do this, open a second terminal, cd into the "DAQInterface" working directory you created earlier, and do the following:
cd $HOME/DAQInterface # Or whatever your DAQInterface working directory is source <products directory>/setup setup artdaq_daqinterface v3_06_00 export DAQINTERFACE_USER_SOURCEFILE=$PWD/user_sourcefile_example source $ARTDAQ_DAQINTERFACE_DIR/source_me
...assuming you haven't put the last three commands in a source-able file as recommended in the previous section. Now that you've set up the environment, assuming you launched DAQInterface in partition 0, run the following:
or, if you launched it in another partition, you'll want to set the DAQINTERFACE_PARTITION_NUMBER environment variable accordingly as well. E.g., if you launched DAQInterface on partition 1, you'd want to do:
export DAQINTERFACE_PARTITION_NUMBER=1 status.sh
Either way, you should see this:
Result: String: 'stopped'
status.shqueries the DAQInterface state machine and prints DAQInterface's state to the screen.
just_do_it.shuses this script to ensure that the transitions it sends DAQInterface went through successfully. We're now in the "stopped" state of DAQInterface - this is the ground state, and means that no artdaq processes exist. In order to launch processes, we send the "boot" transition. To send individual transitions, we use the
send_transition.shscript. However, before we boot the system, we need to specify which components (fragment generators) we want to use. The allowed fragment generators will be listed by DAQInterface if you issue it the following command:
where the output, in the terminal in which DAQInterface was launched, will look something like this:
# of components found in listdaqcomps call: 30 component01 (runs on localhost) component02 (runs on localhost)
..etc. These lines refer to ToySimulator instances designed to run on the current host; in a real experiment you'd expect a more complex listing. They are defined in the file referred to by the DAQINTERFACE_KNOWN_BOARDREADERS_LIST environment variable; in our car this is
known_boardreaders_list_example. Let's select two components (although we could run with just one, or all of them, if we wanted):
setdaqcomps.sh component01 component02
...at the end of the output when you execute this is "Result: Nil"; this is expected, and occurs whenever you send DAQInterface this command, or any of the upcoming transitions. Now we can send the boot transition:
send_transition.sh boot boot.txt
In the terminal running DAQInterface, you'll see something like this:
Fri Mar 8 11:06:59 CST 2019: BOOT transition underway On randomly selected node (localhost), checking that the setup file /home/jcfree/artdaq-demo_subrunrollover_fix/setupARTDAQDEMO doesn't return a nonzero value when sourced...done (2.4 seconds). Launching the artdaq processes Checking that processes are up (check 1 of a max of 5)...found 0 of 6 processes. Checking that processes are up (check 2 of a max of 5)...found 6 of 6 processes. All processes appear to be up artdaq_mfextensions v1_03_03, e15:prof:s64, appears to be available; if windowing is supported on your host you should see the messageviewer window pop up momentarily Determining logfiles associated with the artdaq processes...done (0.0 seconds). Fri Mar 8 11:07:17 CST 2019: BOOT transition complete
If you're sending transitions in the same terminal as DAQInterface is running, hit Enter to return to the command line. The boot.txt file which you edited earlier is used by DAQInterface during the boot transition to determine where to find the script to source before it tries to launch the processes, and then how many EventBuilder, DataLogger and Dispatcher processes to launch and where to launch them. The BoardReader processes which get launched are defined when
setdaqcomps.shis called, combined with the information provided by
known_boardreaders_list_example. What the above output tells us is that DAQInterface has successfully sourced the DAQ setup script on one of the nodes it'll use during the run, that it's successfully launched the artdaq processes, and that since the version of artdaq_mfextensions needed by the MessageViewer application is available, assuming your terminal's display settings are set correctly, a window will pop up which will print messages sent by the artdaq processes. Finally, it determines the names each artdaq process's logfile; this information will be saved later by DAQInterface.
Now that the processes are alive, we'll configure them with FHiCL documents. Simply execute the following:
send_transition.sh config demo
"demo" is the name of the configuration which defines the FHiCL documents used to initialize the artdaq processes. Technically speaking, it refers to the subdirectory which can be found in the
$ARTDAQ_DAQINTERFACE_DIR/simple_test_config directory, and which contains the needed FHiCL documents. Note that you can have multiple arguments to config, each argument being the name of a different subdirectory; in this case each argument is referred to as a "subconfiguration", and by convention the name of the configuration is the set of subconfigurations listed alphabetically separated from one another by a space. If you use subconfigurations, the filenames must be unique across all subconfigurations; e.g., if you use a subconfiguration called "eventbuilders" and a subconfiguration called "dataloggers", you can't have a file called metrics.fcl appear in each subconfiguration- otherwise DAQInterface will throw an exception. It's possible to change the way DAQInterface interprets a (sub)configuration name by editing the user source file, either by switching it to look at another directory on the filesystem or getting it to use an instance of artdaq-database, discussed in the advanced links at the bottom of this wiki. If you wish to see the list of available configurations, just type
and you'll see something like the following:
Available configurations: ascii_simulator_example combined_eb_and_dl config_includes demo demo_largesystem eventbuilder_diskwriting file_closing_example mediumsystem_with_routing_master multiple_art_processes_example multiple_dataloggers request_based_dataflow_example routing_master_example simple_subsystems subconfigs subrun_example See file "/tmp/listconfigs_jcfree.txt" for saved record of the above configurations Please note that for the time being, the optional max_configurations_to_list variable which may be set in /home/jcfree/scratch/DAQInterface_for_updated_documentation/settings_example is only applicable when working with the database
Most of these configurations are intended for developer use; we'll be working with "demo". Scrolling back up to see the end of the output from the config transition you've sent, you'll see the return values of the individual artdaq processes (should be "Success") and then something like this:
Process manager logfiles (if applicable): mu2edaq01.fnal.gov:/tmp/DAQInterface_test_instructions2/logs/pmt/pmt-16799.1-20190204161839.log Mon Feb 4 16:19:18 CST 2019: CONFIG transition complete
Basically, as a courtesy to users - a particularly useful one, if MessageViewer isn't available - the name of the global (i.e., not per-process) logfile for the run is provided, if available. "If available" applies because it's possible to change the method by which DAQInterface launches processes via the DAQINTERFACE_PROCESS_MANAGEMENT_METHOD environment variable (details at the bottom of the wiki), and not all methods support a single logfile. It's possible to watch the logfile update in real time if, in a separate terminal, you type
tail -f <name of logfile>.
Along with the new logfiles, you'll also find in the "/tmp" directory a directory called "run_record_attempted_<USER>/<DATE>", where "<USER>" is your username stored in the environment variable $USER, and <DATE> is the moment the directory of the same name was created. In here, what you'll find are the following:
-The FHiCL documents which are used to initialize the artdaq processes during the config transition
-The DAQInterface boot file, boot.txt
-The metadata file, metadata.txt
-A copy of the script which DAQInterface sourced before launching the artdaq processes, setup.txt
-The list of possible boardreaders DAQInterface used, known_boardreaders_list.txt
-The list of process ranks, ranks.txt. This contains relatively low-level information about the artdaq processes which were actually used in the run, including the host they ran on, the port they listened for commands on, and their "rank". Some diagnostic messages artdaq produces refer to a process's "rank", and this list will translate a rank to the actual process in question.
-The settings file used when DAQInterface was launched, settings.txt
If something goes wrong during configuration, the contents of
/tmp/run_record_attempted_<USER>/<DATE> can potentially be useful for diagnostic purposes. Otherwise, its contents get moved to the run records directory on the start transition, covered next. Click on "Show" for details on the metadata file:
Now let's issue the start transition:
and then DAQInterface will report
Wed Dec 21 16:22:29 CST 2016: START transition complete for run 2
Now we're in the "running" state, as you can confirm if you issue the
status.sh command. Note that the run # is deduced from the run records directory - it's a value one higher than the most recently saved run in the run records directory. If you look in the run records directory you'll now see that the contents of
/tmp/run_record_attempted_<USER>/<DATE> have been copied into the subdirectory called "2/", or whatever the number of the run you just started is.
Stopping datataking and putting DAQInterface in the "ready" state is quite simple:
at which point, you can either start a new run, or you can send the artdaq processes the "shutdown" command and then kill them, via the terminate transition:
You can confirm that you've created a *.root file from the run by looking in the
If you wish to repeat a run you've already taken without needing to worry about remembering which configuration you used, which boardreaders, etc., you can use the just_repeat_run.sh script, like follows:
just_repeat_run.sh <run number> 0
which will use the same boot file, the same configuration and the same boardreaders as were used during the run referred to by
<run number>. Be aware that if there's a mismatch between the code used during <run number> (as determined by the info found in the metadata.txt file for
<run number>and the code which currently exists in the installation referred to by
DAQ setup scriptin the boot file (e.g., commits were made since
<run number>was performed), then just_repeat_run.sh will refuse to proceed; you can override this behavior by passing it the
DAQInterface for your experiment¶
If you've reached this point in the documentation, you're now ready to prepare DAQInterface to work not with artdaq-demo, but with your own experiment's artdaq-based software. You'll want to make copies of the settings file and boot file you've been working with and make the appropriate edits; how to do so is described in The settings file reference and The boot file reference. You'll also want to switch to using the artdaq-database as the source of your configurations rather than the local directory in the DAQInterface package; making this switch is described in Changing the source of the configurations. DAQInterface has a default set of rules as to when to automatically end a run if an artdaq process dies or enters the Error state; to override this set of rules, take a look at Defining which processes are critical to a run.
If you wish to use multiple subsystems in your experiment, the way to inform DAQInterface is described here.To change the underlying method by which DAQInterface launches processes, set the DAQINTERFACE_PROCESS_MANAGEMENT_METHOD variable accordingly to one of the following values:
- pmt means DAQInterface uses the artdaq_mpich_plugin package's pmt.rb script to launch processes. This is the default value.
- direct means DAQInterface directly launches the "boardreader", "eventbuilder", etc. executables from the artdaq package. Practically speaking, the difference between this approach and "pmt" is that with "pmt", if an artdaq process dies when DAQInterface is in the "running" state it ends the run, while with "direct", you have the option of not letting this happen (click on "Defining which processes are critical to a run" a few lines up). Additionally, a feature of pmt.rb is that the output of all the artdaq process is collected in a convenient single log file; this feature is not available in "direct" mode, so each artdaq process's individual logfile must be examined.
- external_run_control means DAQInterface doesn't launch or control processes at all, but instead defers that behavior to some other entity. Currently this option is used by ProtoDUNE. Using it correctly requires providing DAQInterface with information it doesn't typically need; for details contact John Freeman at firstname.lastname@example.org.
Finally, when you call the setdaqcomps.sh command, DAQInterface will look in the file referred to by the DAQINTERFACE_KNOWN_BOARDREADERS_LIST environment variable and use the file's contents to determine the host and (optionally) the port on which to launch the BoardReaders; the corresponding information for the other artdaq process types is provided in the boot file. You'll want to make a copy of the $DAQINTERFACE_KNOWN_BOARDREADERS_LIST file we've been using, and edit it, since you'll want to give your experiment's fragment generators more descriptive names than "component01", "component02", etc. E.g. if you want to use the name tpcreceiver02, then you'll want to both make sure that a file called tpcreceiver02_hw_cfg.fcl exists in the configuration you use, and also that there's a corresponding row in the file $DAQINTERFACE_KNOWN_BOARDREADERS_LIST which looks like this:
where of course you'll want to use a host which is appropriate for your purposes. Also make sure to edit the user_sourcefile_example to set DAQINTERFACE_KNOWN_BOARDREADERS_LIST correctly.
Examining your output¶
Assuming all the transitions went off without a problem¶
The DAQInterface package provides a couple of convenient tools for rapidly examining the results of your run after it's complete. As described earlier, the run records directory provides a great deal of information on the details of the run, but to look for more detail the following scripts can be used.
If your run <run number> was performed with DAQINTERFACE_PROCESS_MANAGEMENT_METHOD set to "pmt", then since all the artdaq process's output gets sent to a single file, you can run this command:
show_logfile_for_run.sh <run number> <open file (0 or 1)>
...will show the logfile for the run number given in the first argument, and will open the file for you (or not) depending on whether the second argument is 1 or 0.
To look at the logfile for a specific artdaq process in your run, you can do the following:
show_logfile_for_process.sh <run number> <process label - e.g., "component01" or "EventBuilder1"> <open file (0 or 1)>
For troubleshooting, you can also do
show_warnings_for_run.sh <run number>
...will show all MessageFacility messages of severity level Warning or Error in the logfile - with the caveat that some warning messages issued by artdaq processes appear in every run, are considered background noise, and as such don't get output by this script, even though they can still be found in the logfiles.
If there was a failure during the config transition¶
In this case, the scripts described above won't work since "run number" isn't a meaningful concept in this case - you weren't able to begin the run at all, and there's nothing in the run records directory. However, DAQInterface does save a temporary directory which contains the files that would have been placed in the run record had the run began; those files can be found in this area: /tmp/run_record_attempted_<your username>/<the partition number you were working on>.
If there was a failure during the boot transition¶
This case is the trickiest to debug, since an artdaq process which doesn't launch doesn't have a logfile. You'll see something like this on the screen, though you may need to scroll up a bit to see it:
The following desired artdaq processes failed to launch:
...followed by the list of processes which didn't launch. Often it's the case that all the processes associated with a certain node are listed, perhaps because of a networking issue (firewall, etc.) One or two things to try: (1) if you're running with DAQINTERFACE_PROCESS_MANAGEMENT_METHOD="direct", on the node(s) where processes failed to launch, take a look at /tmp/launch_attempt_<your username>_partition<partition you were working on>; this file will contain both stdout and stderr from all artdaq processes on the node. (2) trying to boot again, but setting the "debug level" in the boot file to "4", its most verbose level.