Creating and Managing Tasks

The task command

The pscheduler task command is the primary way from the command-line to create new pScheduler tasks. It takes the following form:


The Basics

The simplest task that can be run is a one-shot test with no special options, such as this one which conducts a round-trip time (rtt) test between the local host and, then displays the results:

pscheduler task rtt --dest

Note that the --dest option following rtt is a test option specific to the test type being requested. Each test has its own set of options, so what follows the test type will vary. Most tests that have an option in common that serves the same purpose have consistent names between them. For example, many tests that involve sending data to another host (rtt, trace, throughput, latency) will specify the other hosts using the --dest option.

The task command has a set of its own options that go before the test type. These options are related to how the task is run rather than what test is being conducted. For example, the --quiet switch suppresses the diagnostic information that is usually produced:

pscheduler task --quiet rtt --dest


You can get a full listing of supported task options by running pscheduler task --help. Likewise you can get a full listing of test-specific options by running pscheduler task TEST_TYPE --help where TEST_TYPE is replaced with the type of test you want to run.

Running Tests from Other Hosts

pScheduler determines where to submits a task based on the test parameters. Where a task needs to be submitted is called the lead participant. For many tests run by perfSONAR, a --source switch which specifies where the test should originate and is also the lead participant:

pscheduler task throughput --source host2 --dest host3

If the example command above is run on host1, then the client will submit the task to host2 and the test will be run between host2 and host3.

Not all tests use the source to calculate the lead participant and determining it can get complicated when dealing with things like BWCTL backward compatibility, etc. Luckily, each test plug-in installed on a pScheduler server has the logic required to calculate the lead in the face of this complexity. This does mean though, that the pscheduler command needs to be able to ask a pScheduler server, called the assist server, where a task needs to be submitted.

By default, the pscheduler command will assume there is a pScheduler server running on the local host and try to contact that as the default assist server. If there is NOT a pScheduler server on the local host, then you need to use the --assist flag. For example, say we run the following on host4 and it does not have a pScheduler server but we know that host1 does. Our command could be as follows:

pscheduler task --assist host1 throughput --source host2 --dest host3

The assist server could just as easily be host2 or host3 if they are also running pScheduler servers. It does not matter where the assist server is as long as it a) has a pScheduler server and b) has the test plugin installed for the type of test you want to run.

Selecting Tool(s) for Tasks

By default, pScheduler will automatically select a tool for a task based on three factors:

  1. Availability of tools on each of the participating nodes that can conduct the type of test requested.
  2. The ability of those tools to carry out the test according to the parameters. Not all tools are capable of making good on all of the parameters. For example, traceroute is capable of setting the number of hops allowed by adjusting the time to live but tracepath is not. In a trace test where this is requested, traceroute will list itself as able to participate but tracepath will not.
  3. A preference value hard-wired into each tool by its authors. If multiple tools can carry out a task as requested, pScheduler will select one with the highest preference value or, in the event of a tie, the one that is alphabetically first. The preference value is used to promote the use of newer, better tools while still keeping older, deprecated tools available for situations where they’re required.

Using the --tool switch, the user can control what tool(s) pScheduler selects for a task:

pscheduler task --tool tracepath trace --dest

This will specify that the tracepath tool should be used for running the test. Similarly, repeating the switch forms a list of acceptable tools with those specified earlier being preferred over those specified later. For example:

pscheduler task --tool tracepath --tool paris-traceroute trace --dest

Whether or not tools are specified, a task where no available tool is capable of making the measurement will be rejected.

Repeating Tasks

Any pScheduler task can be configured to run repeatedly by adding options to the task command:

  • --start TIMESTAMP - Run the first iteration of the task at _timestamp_.
  • --repeat DURATION - Repeat runs at intervals of DURATION.
  • --max-runs N - Allow the task to run up to N times.
  • --until TIMESTAMP - Repeat runs of the task until TIMESTAMP.
  • --slip DURATION - Allow the start of each run to be as much as DURATION later than their ideal scheduled time. If the environment variable PSCHEDULER_SLIP is present, its value will be used as a default, and. Failing that, the default will be PT5M. (Note that the slip value also applies to non-repeating tasks.)
  • --sliprand - Randomly choose a timeslot within the allowed slip instead of choosing earliest available

For example, to measure round-trip time 50 times once per hour:

pscheduler task --repeat PT1H --max-runs 50 rtt --dest

It is strongly recommended that repeating tasks apply as much slip as is tolerable to allow pScheduler to work around scheduling conflicts. Larger slip values will will give tasks a better chance of executing. For example:

pscheduler task --repeat PT1H --slip PT30M rtt --dest

Repeating tasks can be stopped using the cancel command.

Archiving Tasks

You can tell the pscheduler command to send results to an archiver using the --archive switch. The definition you give the archiver can take two forms:

  1. A filename starting with the @ symbol that points at a file containing a JSON archiver specification.
  2. A string literal of the JSON archiver specification

For example, the perfsonar-core and perfsonar-toolkit bundles install a special file at /usr/share/pscheduler/psc-archiver-esmond.json with an archiver specification for writing to the locally running esmond instance. You could then use that file to publish a trace test (or any other test) to the local MA instance with the following command:

pscheduler task --archive @/usr/share/pscheduler/psc-archiver-esmond.json trace --dest

Alternatively, you could use a JSON string to accomplish the same as follows (replacing abc123 with the API key used for your esmond instance)

pscheduler task --archive '{"archiver": "esmond","data":{"url":"http://localhost/esmond/perfsonar/archive/","_auth-token": "abc123"}}' trace --dest

For more information on different archivers and their specifications, see Archiver Reference.

Exporting tasks to JSON

The JSON version of a task specification can be sent to the standard output without scheduling using the --export switch:

pscheduler task --export throughput --dest wherever --udp --ip-version 6 > mytask.json

NOTE: Tasks are not validated until submitted for scheduling, so it is possible to export invalid tasks.

Importing tasks from JSON

A JSON file that was previously exported or generated elsewhere can be imported using the --import switch:

pscheduler task --import mytask.json throughput

Test parameters may be changed on the fly by adding them to the command line after the test type:

pscheduler task --import mytask.json throughput --dest somewhere.else

Pausing Tasks

You can pause a pscheduler task so that any scheduled runs will not be executed until the task is resumed. It takes the following form:

pscheduler pause TASK_URL

The TASK_URL is the full URL of the task to be paused and should have been output by the pscheduler task command when a task was submitted. Any runs that would have occurred while in the paused state will be marked as missed. A full example is shown below:

pscheduler pause

Resuming Tasks

You can resume a previously paused pscheduler task so that its runs will again be executed. It takes the following form:

pscheduler resume TASK_URL

The TASK_URL is the full URL of the task to be resumed and should have been output by the pscheduler task command when a task was submitted. Future runs will be carried-out normally after being resumed. A full example is shown below:

pscheduler resume

Canceling Tasks

You may cancel a task with the pscheduler cancel command which takes the following form:

pscheduler cancel TASK_URL

The TASK_URL is the full URL of the task to be canceled and should have been output by the pscheduler task command when a task was submitted. This command cancels any future runs of the task specified. Any run of the task which is underway will continue to completion. The task will still be in the database but will be marked as disabled. This means you will still be able to query results of runs completed prior to cancellation but no new results will be generated. A full example of the command is shown below:

pscheduler cancel