We see heaps of CRC/Alignment errors in our switch logs, but they don't seem to have a cause.
We have just fired up 10 brand new 3424, 3448 and 6400 switches, all connected by 10Gb copper connections. The switches are meshed, using the 10Gb ports.
We get CRC/Alignments errors on all 10Gb links where data is passed. Every 5 minutes when they are actively used, not so often when they are idle (ie. just sending internal, administrative mesh traffic).
The only host traffic passing through these switches is to 1 server, the only one plugged into these switches, on a 3448. It downloads data from other servers, that are on a Nortel-switch, connected to the ProCurves via a 4x100Mb trunk on a 3424.
|Nortel| -> Trunk -> |3424 in mesh|
HPs own techies and support people cannot explain the messages.
So, there is virtually no traffic going through anything here (max has been 200mbps), all configs have been checked, firmware updated, you name it.... still we seem to have "excessive CRC errors". There are no glitches in data transfers either, so we see no problems at all.
I suspect the reporting of the errors is flawed, that the number of errors is not "excessive" at all.
It can't be a cable issue, or else all our 10Gb cables are broken. Nor can it be a speed setting issue, because we can't configure speed for the 10Gb links.
Anyone seen anything like this? I'm giving up ... Cheers, Anders RM :)
>It can't be a cable issue, or else all our 10Gb cables are broken. CRC errors are generally a hardware (port) or cabling issue. Now... I have not worked with 10 gig over copper so cannot say for certain that it is less tolerant of off-spec cabling but think it is a safe assumption.
Not sure of your reasoning about "else all... are broken".
Excessive CRC/alignment errors this Alter Log Message is appearing because of A high percentage of data errors has been detected on this port. Possible causes include: â ¢ Faulty cabling or invalid topology. â ¢ Duplex mismatch (full-duplex configured on one end the other) â ¢ A malfunctioning NIC, NIC driver, or transceiver
which nortell switch do you use in your topology, check is it support meshing
Thanks for the answers, guys. Let me clarify a couple of things.
Les: >>It can't be a cable issue, or else all our >>10Gb cables are broken.
>I have not worked with 10 gig over copper >so cannot say for certain that it is less >tolerant of off-spec cabling but think it >is a safe assumption.
I have been told by HP that this cabling is very sensitive. The errors have occurred on /all/ meshed ports at some point. As far as I can tell, usually when we see "real" data going across the links (there is a lot of intra-mesh babbling on these links). Hence, if this is a cabling issue, all cables must broken. Also, changing cables on some links did not change the situation at all.
Nor did changing the interface cards (10Gb ports) for which we received the most errors.
I find it highly unlikely that all of our ports/all of our switches have the same error.
Mohammed:
We went through all the rules of meshing, and all of the setup (basically there is no setup... it is all plug and play so far.)
Duplex mismatch is ruled out, since there is no such setting for 10Gbps.
Faulty transceivers? Well. in that case, all 13 cards we have tried so far are faulty.
Invalid topology is the only option that I can't really rule out myself, but our topology was designed and set up by an HP ProCurve specialist, confirmed to be OK by HP support when we reported the issue, and as far as I can tell from documentation, it is correct. HP has been helpful in trying to track down the error, but with no luck so far.
The Nortel is not part of the mesh, and we experience no errors on the link (trunk) to it.
The weird part is that we have no problems getting data through. So, I'm still leaning towards faulty diagnostics.. I don't think there actually is an issue, but I'd like to have that confirmed before I start depending too heavily on these switches. Also, if that is correct, I'd like to see HP get to work on a bugfix...
