perfSONAR User Q&A and Other Discussion

[Pete LG <>]

Hey Andrew, It's been a while. Thanks for getting this patch deployed!  I setup some dev perfSONAR vms to test the patch in the psconfig scheduling bug.

Maddash still seems to be on a different schedule though more than once a week.

Here's the config line you sent me while back earlier in this thread.

"schedules" : {

      "schedule_once_a_week" : {

         "repeat" : "P1W",

         "slip" : "PT12H",

         "sliprand" : true

      }

}

This is from a single specific "check detials" in Maddash. All the sites with this schedule show this:

At the bottom of the check for specific throughput test, it says this:

---

Check Interval	This check is run every 1800 seconds when there is no change in the result from the previous run
Retry Attempts	A result must be seen 1 times before changing the status.
Retry Interval When a change is observed the check will be run every 300 seconds -- Why would it keep checking every 1800 seconds, if it doesn't update more than once a week? Any help will be appreciated.  Thanks again!

Pete LG

Sent with ProtonMail Secure Email.

‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐

On Tuesday, March 5, 2019 2:01 PM, Pete LG <> wrote:

That's great, thank you so much.  The support you guys provide is simply amazing :)

Thanks!

Pete LG

Sent with ProtonMail Secure Email.

‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐

On Tuesday, March 5, 2019 1:45 PM, Andrew Lake <> wrote:

Hi Pete,

Thanks, I think you found a bug with respect to the test running too often. I created an issue here https://github.com/perfsonar/psconfig/issues/59. You are correctly specifying the 1 week interval, but pSConfig injects a separate “start” and “until” time that indicate how long the task is good. The default for this is 24 hours. It does this so tasks don’t get orphaned and if something goes wrong a task doesn’t get stuck on a server forever. The problem is that it is not compensating for intervals over 24 hours, which it should. It’s actually a pretty straight-forward fix I think, so we will try to get something in the next release. 

As for the IPv4 vs IPv6 issue, the two endpoints of the test you shared are specified as DNS names both of which have AAAA records. If not otherwise specified, the test will default to using IPv6 since it is available on both ends. This is the intended behavior. If you want to force it to use IPv4 in such a case then you can set the “ip-version” option to 4 in your test or specify at least one of the addresses as an IPv4 address. 

Hope that helps,

Andy

On March 5, 2019 at 1:34:03 PM, Pete LG () wrote:

Hey Andrew, Wow, that was quick!

If you look here at the TPut graph, it seems to have several check going
on throughout the week:


https://psm02cou.col.more.net/maddash-webui/details.cgi?uri=/maddash/grids/MOREnet+Mesh+-+TCP+Throughput+Test+Between+MOREnet+Bandwidth+Hosts+-+Throughput/co-core-psr.mo.more.net/sl-dist-psr.mo.more.net/Throughput

I'm using the schedule you provided below to do it:

"schedules" : {
      "schedule_once_a_week" : {
         "repeat" : "P1W",
         "slip" : "PT12H",
         "sliprand" : true
      }
}

Also, do you know if this is doing IPv4 also?  I thought if you didn't
specify it did both, but I can only see IPv6 here. (At least, I believe
the old version worked like that.


Here is the config:

https://psm02cou.col.more.net/psconfig/ptest.json


Thoughts?

Thanks!!

Pete LG

Sent with ProtonMail Secure Email.

‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐

On Tuesday, March 5, 2019 10:08 AM, Andrew Lake <> wrote:

Hi,

Are you asking because MaDDash seems to be checking for a result more than once a week or because your graphs show more than one result per week? MaDDash is just the messenger, it doesn’t execute any tests it just periodically asks the archives for results. You can control how often it checks independent of how often the test actually runs and having MaDDash query the archive more frequently is probably a good idea so you know if anything goes down more quickly than you would otherwise. 

If the test seems to be running more often than a week, how much more often? If you look at the JSON where you define the test are you using any slip and.or do you have sliprand set? That can cause some variation. 

Thanks,

Andy

On March 5, 2019 at 10:24:06 AM, Pete LG () wrote:

Hi Andrew, I set this up several weeks ago per your recommendation, but I noticed from the graphs that there seem to be more than one check during the week.

Does Maddash strictly follow the schedule you prescribed, or is there more going on.

I'm happy with it, but I'm curious what exactly is going on.

Thanks!

Pete LG

Sent with ProtonMail Secure Email.

‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐

On Wednesday, February 6, 2019 8:28 AM, Andrew Lake <> wrote:

Hi,

All the docs for defining a mesh, now part of the pSConfig component, can be found under the heading here: http://docs.perfsonar.net/#managing-multiple-hosts-with-psconfig. It goes into quite a bit of depth on concepts as well as references for the syntax.

Examples can be found under /usr/share/doc/perfsonar/psconfig or you can view them online here: https://github.com/perfsonar/psconfig/tree/master/doc

If you want a really large example, you can see ESnet’s pSConfig template that drives ps-dashboard.es.net here: https://github.com/esnet/esnet-perfsonar-mesh/blob/master/psconfig/esnet-psconfig.json

For running a test once a week, update your psconfig templates “schedules” section with something like the following:

"schedules" : {

      "schedule_once_a_week" : {

         "repeat" : "P1W",

         "slip" : "PT12H",

         "sliprand" : true

      }

}

The “repeat" says you want it to run every week, the “slip” gives it 12 hours of wiggle room if it cannot get a slot exactly 1 week from the previous run, and “sliprand" has it randomly choose a slot in the 12 hour range so everything is not piled at the beginning of the interval. The 12H is arbitrary and probably a bit high, but its good to build in some wiggle room, especially if you have a lot of tests. 

Thanks,

Andy

On February 5, 2019 at 9:14:09 PM, Pete LG () wrote:

I have a request to add a maddash/mesh grid for 'throughput' testing that is only run once a week at a certain time. 

Does anyone know the best way to go about this.  I suggested that we cron off a test and just get an email once a week with iperf results, but they asked for it in maddash form.

Any ideas or suggestions are most welcome.

Also, is there any work being done on the newer docs for maddash/mesh?  From what I can tell, most of the examples are still in yaml, and all the good examples show bwctl versions of tests.  If anyone knows of example out there, I'd love to view them.

Thanks,

Pete

Pete LG

Sent with ProtonMail Secure Email.

--

To unsubscribe from this list: https://lists.internet2.edu/sympa/signoff/perfsonar-user