@Erfan7 wrote:
Joseph my question is why devices were fine with old switch, and why some didn't come up with new switch, as from device end nobody touched the cables/connectors only cables were reconnected to new switch in IDF, when connectors were made from device end they came up, let me tell you this was happened many times, and I also know issue was with connectors as you said internal connection was marginal.
Two possibilities come to mind.
First possibility, the link, end-to-end, never met spec, due to one or both RJ45 connectors not meeting spec, and the replacement switch isn't as tolerant of the link being out of spec as the earlier switch was.
If this is happening, why the two switches don't use the same Ethernet tolerances is a fair question, but that's a whole different line of inquiry vs. that they do.
Again, if the newer switch isn't as tolerant, the reconnected link fails, even though, it (the link), hasn't changed at all.
Thinking about this possibility, it might be the most likely, even over any possible issues caused by disconnecting and reconnecting the switch end RJ45. It can also explain why either end, or both ends, RJ45, not meeting spec, would be an issue on the replacement switch.
The second possibility, is a bit more exotic. So, you pull the switch end RJ45, swap switches, insert the switch end RJ45, and have a connectivity issue. Then what do you do? I.e. how did you "find" replacing the far side RJ45 to be a fix? I would think, your first action would be to first replace the switch side RJ45, and only if that did not fix the issue, then try replacing the far side RJ45. So, if that's what you've been doing, it would imply, the switch end RJ45 wasn't the problem, only the far side RJ45, but that side, hasn't been touched yet, so it couldn't be the problem, right?
Well, the forgoing assumes the replaced switch side RJ45 performs as well as or better than the RJ45 it replaced. But if it didn't at least match the replaced RJ45 performance, and the far side is sub marginal, the combination of replaced near side, and far side sub marginal RJ45, precludes link working until far side is replaced with a "better" performing RJ45, whose improvement, is enough to allow end-to-end connectivity.
The only way I can see narrowing down this issue, to root cause, would be to use an end-to-end cable tester and/or after a reconnected RJ45, switch side, link failure, reconnect to prior switch (the latter not possible if being replaced after actual switch failure, not just for switch replacement for a newer switch).
BTW, hardware device tolerances are sometimes better than what a spec allows for, but usually no one complains about being able to exceed spec, until you bump into a case where suddenly meeting spec becomes a problem.
For example, we all know the copper Ethernet specs usually calls for a 100 meter distance limitation, but have you ever tried running 101 meters? Do you think 101 meters will definitely not work, or there's some chance it will? If some chance it will, what about 110 meters, 120 meters, 150 meters? What about the "quality" of the cable itself? Do you think some cables might just go a bit longer than 100 meters then other cables? How/why is there a 100 meter limitation, how does a host know or detect cable is too long? Could different host NICs have different allowance for how far 100 meters might be exceeded?
Per chance, were older switches using 10/100 and newer 10/100/1000 ports? What speeds are you running at, and what Cat level cable is being used?
Can you tell us, that before a failed switch was replaced, its ports show zero L1 errors?
In other words, good chance, as @liviu.gheorghe notes in a recent reply, your cabling plant is "poor quality" and if so, and if newer switches are less tolerant, you would see the continuous issues as you replace switches, for whatever reason.
If the problem often is resolved by replacing RJ45 connectors, then, most likely, most might be sub spec.
Again, as I noted earlier, things like device vibrations and/or manually manipulating cables, might impact their operational capability, but this, and newer switch tolerance differences, aren't mutually exclusive.
