perfSONAR User Q&A and Other Discussion

["Ahuja, Satish (Exchange)" <>]

Hi Mark,
Hope you find this email well. I re-built the node1 and it’s working as expected without any errors. I can also see some grids on MaDDash for previous tests, but I see one error that says grid is down.

 

Unfortunately, node two is throwing some error, I tried rebooting it but it persists.

Will you recommend re-building node2 as well? Thanks

 

Regards

Satish Ahuja

Student Systems Administrator

OIT NEO, CU Boulder

 

From: Mark Feit <>
Date: Friday, February 24, 2023 at 10:28 AM
To: Ahuja, Satish (Exchange) <>
Cc: Conan F Moore <>, Satish Ahuja <>, <>
Subject: Re: [perfsonar-user] perfSONAR User Q&A and Other Discussion

Ahuja, Satish (Exchange) writes:

 

I ran the troubleshooting steps you mentioned and found error on perfSonar node1. The process or daemon is unable to connect to database. The other node is working fine. I wonder I followed the same steps while installing image but somehow node1 got into some errors. I will try to troubleshoot this otherwise I will start over on node1. I will keep you updated once I’m done with this.

 

Based on what you sent, it looks like PostgreSQL, which underpins both pScheduler and Esmond, didn’t come up properly on that system.  We’d have to do a lot of forensics to figure out why and would ultimately end up doing a forced reinstallation of the affected packages.  I’d recommend doing that, which involves running these two commands as root (assuming you’re on CentOS and not Debian or Ubuntu):

 

• yum -y erase ‘postgresql*’  (This will erase a lot of packags.  Do not panic.)
• yum -y install perfsonar-toolkit

 

That should put the system back in order and into its unconfigured state.  If it doesn’t work after that step and setting pSConfig to grab your mesh file, something isn’t right on that system and I’d recommend nuking it from orbit and rebuilding the system from the OS.

 

--Mark