Maybe a Potential Issue when modeling the module bays the new way

[KaricAllegra]

Ok... So this is gonna be tough. I am in NO WAY anything more than a glorified Netbox GUI user, so much of what I am going to try and explain here is probably NOT going to help anyone figure it out, BUT I want to at least get it out in front of some smarter people to hopefully see if my issues are just a one-off.

Oh, and making this MORE fun: my Netbox instance is air-gapped, and all this was on my work computer, so I can't even copy and paste all the error lines; I have to try and remember them to relay them to you smarter folk. So again... apologies up front. Here goes:

My issues started about 1 week ago. I have been setting up sites for a customer, with the first building completed about 3 weeks ago. I followed the newer Netbox guidance and created EVERY device that should have a module bay, with a module bay. So all of our Corning Edge modular fiber patch panels were modeled as either a 1RU device with 12 Module bays or a 4RU device with 48 Module bays, and so on. Additionally, all our Cisco switches were modeled to include a module bay for the Network module, and, for the fiber switches, an additional 48 module bays for transceivers. Thanks to this, we were able to track all our Corning cassette modules, Cisco network modules, etc., and it has been GREAT.

I began the second building at our Site Group, and that is when the issues seemed to start. I do not know if this is a coincidence or related to the start of imports for the 2nd bldg. Now, when I attempt a bulk import for a Telcomm room (say, an average of 50 devices), I get a timeout, and nothing gets imported. It just doesn't seem to work unless I do like 2 devices, which, of course, is untenable. So, I reached out to the old Google Gemini Pro to see if it could help me. (Please again, realize I am NOT a Linux guy lol)

In going down that AI rabbit hole, the issue seems to be tied to something in how Postgres has to recalculate all the module positions each time you attempt to import new modules. Since one of the CLI responses seemed to indicate it was a tree calculation issue. AI said it was because each of the 84 devices I have, with 48 module bays, plus the 200+ Cisco switches with 49 module bays, was now having trouble calculating “where” they were in a database hash tree. Again, this is AI's take, based on the commands it was telling me to run, from a guy that learned what HTOP was during this exercise lol. What I for sure remember from the issues is that running an HTOP after trying to import would show a Postgres PID that was maxing out a core, and would often run for like 15 min before failing. It ended in the word UPDATE.

So again, I am so sorry that I am not able to just have a bunch of those logs here for smarter people to look into, but I am REALLY hoping someone will see this and be like, “man, you just need to do X” cause dropping 2 lines at a time is going to take FOREVER.

Further info, in case it may be of T/S help:

• I have upgraded NetBox 3 times now to see if it was maybe an issue that was already caught.

• Yes, I have restarted all the services and even rebooted both the VM and the Hypervisor.

• No, I can’t print out logs to give anyone, BUT if someone has some commands they think could help isolate the issue, I could run them and at least tell you what happens.

• No, I probably can’t give you much more detailed info because, you know - government.

So, in closing, anyone who takes the time to read all this and offer ANY help, if you are ever in the DC area, any drinks/food you want are free at my bar, which is located in Leonardtown, MD (bout an hour south of DC)

[KaricAllegra]

This issue seems to have been fixed by #22146

So if you are seeing this performance hit, upgrade to at least v4.6.1