Sun E450 Incrementing Errors Attached to 3750

Unanswered Question
Dec 23rd, 2008
User Badges:

I have been revamping the network for one of our departments. I installed the new equipment, a stack of 2 3750G-48-TS's. Now, when attached to an SX SFP, a Sun E450 server (around 9 years old; it has a GBIC-style SC connector) is incrementing input errors.


The server used to be on a 4003 using an SX GBIC. There were never any reports of errors with it, but the server has been rebooted multiple times and I have no way of knowing for sure. During our diagnostics, we connected it to a 3550 with an SX GBIC. The server does not increment errors on that.


When attached to the SFP on the 3750, the server gets input errors. There are no errors on the switch side. We have tried different SFPs. We have tried different fiber cables. We have put the server on a different port on the same stack. We put a different server on the same port that the E450 was on. It did not increment any errors. We tried swapping interface cards on the server and get the same errors. But when we put the server back on a 3550, the errors stop.


At this point, my thinking is that there is some background protocol that is going on that the E450 doesn't recognize. Both the 3550 and 3750 switch ports are configured identically (access mode, static VLAN assignment). The port is not a trunk so it's not VTP (which is set to transparent on both). The 3550 didn't have UDLD turned on, so I tried turning it off on the 3750 and that didn't fix it. MDIX doesn't apply to fiber interfaces. I did a side-by-side comparison between the 3550's configuration and the 3750 and I don't see any other differences that would account for the input errors.


The fiber cable we are using has been run through a Fluke tester and tested good.


The server's communication is symptomatic; there are delays and slowness complaints trying to communicate with the server, so there is definitely some kind of communication problem. These problems go away on the 3550.


The only other thing I can think of if not some kind of unrecognized background protocol would be a slight variation over time in the implementation of the gigabit encoding (GBIC vs SFP) such that the server is being sensitive to it. Again, this machine is 9 years old. But it doesn't make any sense that we can plug the server into the 3750, get errors, unplug it and plug in the cable to the 3550 and the errors go away.


These devices are all in the same room with cables no longer than 30'.


Ideas??


Thanks,

Matt

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.

Actions

This Discussion