Question for network engineers regarding link fail over.
I will preface this by saying I am not a network engineer however I do mess around with a lot of networking equipment in my home lab.
By trade I am a systems engineer.
In one of the data centers that I work on we had a OSPF routing issue that the network engineer determined was due to a flapping interface.
In this design there is a distribution switch that is dual homed to two disparate core routers. Each connection from the distribution switch is a single link. The core routers have a single link between them.
Yesterday started getting alerts that some of our servers were unavailable. However they were up it was just an issue of connectivity between specific subnets in the network.
Our network engineer diagnosed the issue as a flapping link to core router one. According to him this is the shortest path via OSPF. However the link was flapping a lot so OSPF was apparently in a bad state as routing tables constantly trying to determine the correct path and whether to route packets to core router one or core router two.
I asked why do we not have two connections to each core router from the distribution switch. Port channels is the terminology that I am familiar with from messing around with Cisco but these are juniper devices so I think it's called something different.
The network engineer told me that if we had LACP connections to each core router the issue would have actually been worse. As one of the links to router a would have been fine and one of the links to router would flap.
Honestly I don't know enough about networking to prove this person wrong. However I've worked in Enterprise data centers enough and I'm with enough Enterprise data center network engineers to know that they have redundant links everywhere.
I'm trying to figure out if this network engineer just doesn't know how to do what I am asking him to do.
I want redundancy everywhere, redundant links between all devices, and redundant paths such as a distribution switch having the ability to route to two different core routers.
If I go to Google Gemini or chat GPT and provide the network engineer feedback, I'm always told it's just an AI hallucination.
Beyond redundant links to each core router. I would think that there should be some sort of OSPF configuration that would blacklist the flapping connection. If we were using whatever juniper's equivalent is for ether channel, I would think that ideally we'd never even have to have a routing table change and we would just know that we have a singular bad link and we need to troubleshoot transceivers and/or cables.
I'm also not 100% sure that I believe that the constant flapping of the interface was the core issue. My understanding is that OSPF converges pretty fast and that even if it link was flapping that we wouldn't lose connectivity as long as both of the northbound core routers or configured properly.
Can someone provide me some insight here, and possibly make some suggestions to how I should approach better resiliency and redundancy to the network engineer.
Sorry for the long post I didn't know how to make it much shorter than I did.
