Skip to Content.
Sympa Menu

perfsonar-user - Re: [perfsonar-user] Oddities with the Alama9 install of pst5.0.8

Subject: perfSONAR User Q&A and Other Discussion

List archive

Re: [perfsonar-user] Oddities with the Alama9 install of pst5.0.8


Chronological Thread 
  • From: "Robinson, John-Paul" <>
  • To: Andrew Lake <>
  • Cc: "" <>
  • Subject: Re: [perfsonar-user] Oddities with the Alama9 install of pst5.0.8
  • Date: Thu, 11 Apr 2024 14:57:11 +0000
  • Msip_labels:

Hi Andrew,

Thanks for the comments.

I don't mind configuring NTP from the backend.  The OS takes care of reasonable defaults, so I trust that.  I just like to see my web status of NTP say "yes" and not "no". 😉

Also, glad to hear of the SELinux updates.  My preference is for the Ubuntu release.  I just went with Alma to try out something in the post CentOS era and because Ubuntu 20.04 has less runway.

Regarding the bi-directional tests, I am able to run them just fine.  The problem is that the results aren't showing up in my graphs.   I checked the log file you mentioned and see errors like the following, which I'm guessing points at the problem:

2024/04/11 08:11:38 WARN pid=2393 prog=perfSONAR_PS::PSConfig::PScheduler::Agent::_run_end line=229 guid=BFC42B2C-F804-11EE-8119-206FAD4DA91F msg=Problem deleting test throughput/iperf3(ps40g.iu.xsede.org->138.26.220.66), continuing with rest of config: FORBIDDEN: Forbidden.

2024/04/11 08:11:38 WARN pid=2393 prog=perfSONAR_PS::PSConfig::PScheduler::Agent::_run_end line=229 guid=BFC42B2C-F804-11EE-8119-206FAD4DA91F msg=Problem adding test throughput(138.26.220.66->cinc.ps.internet2.edu), continuing with rest of config: INTERNAL SERVER ERROR: Error getting tools from cinc.ps.internet2.edu: Failed to connect to cinc.ps.internet2.edu port 443: Connection timed out

2024/04/11 08:11:38 WARN pid=2393 prog=perfSONAR_PS::PSConfig::PScheduler::Agent::_run_end line=229 guid=BFC42B2C-F804-11EE-8119-206FAD4DA91F msg=Problem adding test throughput(cinc.ps.internet2.edu->138.26.220.66), continuing with rest of config: Network is unreachable

I'm guessing these combine to describe why the record of the tests is not getting logged in my graphs, I just don't understand why this would be happening.  I have "bi directional" selected in the test config and in prior releases I haven't had to do anything extra beyond that to get the bi directional results to show up in my graphs.

Thanks for any pointers.

John-Paul


From: Andrew Lake <>
Sent: Wednesday, April 10, 2024 2:53 AM
To: Robinson, John-Paul <>
Cc:  <>
Subject: Re: [perfsonar-user] Oddities with the Alama9 install of pst5.0.8
 
Hello John-Paul, 

A few comments:

- We don’t support the web ui to edit chronyd on EL9, so it will not be shown. For security reasons, we are moving away from editing system files via the web UI like that, so no plans to add support in future. There are some notes in the docs, but they are easy to miss. 
- We are super-close to a 5.1.0 beta release (RPMs uploaded, just sorting some issues out with our Debian repo) which has a bunch of cleanup for SELinux which should help with the issues you encountered,
- You are correct that there is not way to import tests from 4.X


WRT bidirectional issue, are you able to run a reverse test from the command-line? For example:

pscheduler task throughput --source atla.ps.internet2.edu --dest ps-sd.rc.uab.edu

The above worked for me as a third-party test, so I suspect it will work for you too but would be good to check. Assuming that works, any complaints in /var/log/perfsonar/psconfig-pscheduler-agent.log?


Thanks,
Andy

On April 9, 2024 at 10:19:57 PM, Robinson, John-Paul () wrote:

I confirmed the chronyd service was running using systemd 

systemctl status chronyd

I wasn't able to control it via the web or see it reflected on the web page as ntp running. 

The thing I haven't figured out yet is getting bi directional tests to run.   I only see performance data for test results from my site to remote sites, not the reverse. I think it could be an issue with recording the results more so than running the tests. Not sure at this point and haven't gotten back to it for a couple of weeks to debug further. 

From: Ricard Cruz <>
Sent: Tuesday, April 9, 2024 8:56:00 AM
To: Robinson, John-Paul <>
Cc:  <>
Subject: Re: [perfsonar-user] Oddities with the Alama9 install of pst5.0.8
 
You don't often get email from . Learn why this is important
Hello John-Paul,

Thank you for sharing your experience with the reinstallation of these nodes on Alma9.

We are currently undergoing the same process, and I would like to inquire if you ultimately managed to configure NTP/Chrony via the web interface.

Thanks in advance.

Best regards,
Ricard

El mié, 20 mar 2024 a las 0:47, Robinson, John-Paul (<>) escribió:
As another update, I was able to resolve the "Error retrieving data No data found in the measurement archive" when clicking on throughput test result link from the main web page by setting SELinux to permissive and then rebooting.  After the reboot the plots were correctly rendered, and the historical data displayed (data collected while previously getting the above error). 

I'm still not seeing the bi-directional test results. Just one-way, outbound from our perfsonar node.   I'll observe if this changes as a result of the permissive change.

The selinux setting change also fixed my ability to alter the lsregistrationdaemon.conf file via edits from the web page.  I am now able to make changes there and have them persist, instead of resetting the form to blank after clicking save.

John-Paul

From: Robinson, John-Paul <>
Sent: Friday, March 15, 2024 9:58 AM
To:  <>
Subject: Re: Oddities with the Alama9 install of pst5.0.8
 
By way of an update, I was able to fix the lsregistration configuration by manually editing the config file.  It seems that the status on the toolkit web page that our site wasn't global registering was indeed accurate.   The root cause seems to have been the auto discovery of interfaces and addresses.   Our ConnectX card was on interface enp3s0.  It was either the fact that this was an undetected interface or that Alma9 has moved away from the tradition ifcfg based interface configuretions that prevented the site from being able to discover it's IP address.

By turning off autodiscover_interfaces and adding the following stanza, I was able to avoid the lsregistration start up errors and warnings and get a successful registration.  The toolkit web page then showed the site was registered.

        autodiscover_interfaces    0
        primary_interface       enp3s0
        host_name ps-sd.rc.uab.edu
        address 138.26.220.66
        mtu 9000
        capacity 40000

        <interface>
             if_name enp3s0
        </interface>

I don't think the mtu and capacity are strictly necessary but the Perl code was throwing warnings during startup that the values were uninitialized, ie. it couldn't discover them from the host.   I added them just to avoid any non-INFO level mesages during startup.

I'm still debugging the bi directional tests and test archive issues.

From: Robinson, John-Paul <>
Sent: Thursday, March 14, 2024 11:05 AM
To:  <>
Subject: Oddities with the Alama9 install of pst5.0.8
 
Hi Folks,

We reinstalled our perfsonar node using Alama9 and following the perfsonar toolkit install here with straight cut-n-paste.


I've seen some oddities on the install compared to our prior install (4.4 on centos6). 

  • I haven't been able to update the host information configuration via the web UI.  I can fill out the form, but when I press save it just resets to blank. 
  • I worked around this by editing the underlying config file /etc/perfsonar/lsregistrationdaemon.conf directly.
  • Unfortunately, I can't seem to set the organization name.  There isn't a field for it in the file and I haven't guessed the correct one yet.
The site overview says my NTP config is not active and not registered
  • I'm hoping these are UI inconsistencies: chronyd is running and so is the perfsonar-lsregistration service.
I noticed I couldn't import my old test data, because it seems Cassandra is not used.
  • That's OK, just confirming my understanding.
I defined new tests and they are running but I only see performance data in one direction (the tests show bidirectional in the web UI) and, more disturbingly, when I click on the graphs there is an error "Error retrieving data No data found in the measurement archive" instead of a graph.
  • I've checked all my running processes and they are all active except for "perfsonar-generate_motd.service" and "perfsonar-configure_nic_parameters.service" which exited.
  • looks like the nic config is attempting to use old style ifcfg changes "perfsonar-configure_nic_parameters[1294]: ls: cannot access '/etc/sysconfig/network-scripts/ifcfg"

Here's our site for reference https://ps-sd.rc.uab.edu/toolkit/

Thanks for any insights you can provide on the above.

John-Paul

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


--
-----------------------------------------------------------------------
Ricard Cruz
Port d'Informació Científica (PIC)              e_mail:
Campus UAB - Edificio D                         Tel: +34 93 170 27 28
08193 Bellaterra (Barcelona), Spain      
http://www.pic.es
Avis - Aviso - Legal Notice: http://www.ifae.es/legal.html
       
-------------------------------------------------------------------------
--
To unsubscribe from this list: https://lists.internet2.edu/sympa/signoff/perfsonar-user



Archive powered by MHonArc 2.6.24.

Top of Page