reddragdiva: (Default)
[personal profile] reddragdiva

Anyone who herds lots of Sun kit knows the one where the Sun box and the Cisco router fail to autonegotiate. "After you." "No, after you." "No, I insist." "No, no, no." Hence locking both to 100 Mbit/full duplex (or 1Gbit/full) as the only way to get them to behave.

What I want to know is why this bug hasn't been sorted out in years and still exists in 2005. It's not like it's a rare combination.

(no subject)

Date: 2005-05-18 11:29 am (UTC)
From: [identity profile] simonb.livejournal.com
Its interface device specific; hme still suffers from it & my gut reaction is that its thanks to hardware issues in the network device itself.

I've not had any problems with the dmfe, (Netra X1, SunFire V100), ce (PCI GigE card and others) or bge (SunFire V210/V240) interfaces. On modern kit I have occasionally seen it surface on the eri interfaces (SunFire V120, SunBlade 100) - that was with them being connected to HP ProCurve switches as well.

(no subject)

Date: 2005-05-18 11:31 am (UTC)
From: [identity profile] gmh.livejournal.com
The really fun bit is where the Cisco kit fails to autonegotiate properly with ... the Cisco kit.

I look after 7206s, 5500s, 2950s and sundry Pixes. Every single one of them has occasional autonegotiation issues with one or more of the others.

At which point you just set everything to manual 100full.

(Then you only get autonegotiation problems with Win2k3. Which is much less hassle.)

G.

(no subject)

Date: 2005-05-18 11:35 am (UTC)
From: [identity profile] rbarclay.livejournal.com
I remember a Cisco rep proudly presenting the 2600 series a couple years back. "And, finally, autonegotiation works with every other kit we could get our hands on for testing." he said, with a big fat smile.

I just connected one 2620 to another, and foom, no autoneg, just as always.

(no subject)

Date: 2005-05-18 11:36 am (UTC)
From: [identity profile] leadgend.livejournal.com
Because you have a workaround so the bean counters don't want to spend anything on fixing it?

(no subject)

Date: 2005-05-18 12:22 pm (UTC)
From: [identity profile] pir.livejournal.com
It has been fixed as far as I can see. I haven't had any autonegotiation problems with sparcs in many years other than connecting them to very very old catalysts. When we upgraded the catalysts, problem went away.

(no subject)

Date: 2005-05-18 12:24 pm (UTC)
vatine: Generated with some CL code and a hand-designed blackletter font (Default)
From: [personal profile] vatine
Because auto-negotiation s[pecs are not as bomb-proof as they could be? Not sure that *is* the case, but it certainly would not surprise me. Hard-coding is good, though, keeps you on your toes. I wonder when kit will start coming with thermite charges integrated, to be ignited on syntax errors?

(no subject)

Date: 2005-05-18 12:28 pm (UTC)
From: [identity profile] dan-lane.livejournal.com
Because a Sun engineer called up a Cisco engineer and proclaimed "hey, we have this problem!,.. you want to fix it?" ... to which the conversation quickly deteriorated... "After you." "No, after you." "No, I insist." "No, no, no."

(no subject)

Date: 2005-05-18 12:42 pm (UTC)
From: [identity profile] grumpy-sysadmin.livejournal.com
Best. Explanation. Ever.

(You'd get bonus points if you had explained that Sun and Cisco had to find an alternate communications channel--say, email--to make their digital phone systems properly connect, prefaced by lots of rings that, when picked up, were dial tones.)

(no subject)

Date: 2005-05-18 12:44 pm (UTC)
From: [identity profile] bellinghman.livejournal.com
Which part of 'other' didn't you understand?

*snigger*

(no subject)

Date: 2005-05-18 03:06 pm (UTC)
From: [identity profile] hellsop.livejournal.com
Exactly. Everyone that cares learns "autoneg doesn't work", and starts routinely locking protocols on that suite of kit, and never runs into the problem again. Even if they file a bug report, it's never above a "annoyance" severity, because now they're working, and anyone else that has the same problem can also fix it the same way for free.

(no subject)

Date: 2005-05-18 04:00 pm (UTC)
From: [identity profile] sweh.livejournal.com
Which fails to work for network jumpstarting, of course, since the devices don't negotiate well enough to talk to the network to get to the boot server! Grump.

(no subject)

Date: 2005-05-18 04:48 pm (UTC)
From: [identity profile] rbarclay.livejournal.com
I always rather imagined it as: the two engineers met over a beer, and concluded that the first to listen to the lusers whining would have to pay the next round...

(no subject)

Date: 2005-05-18 05:49 pm (UTC)
ext_8707: Taken in front of Carnegie Hall (LISA `97)
From: [identity profile] ronebofh.livejournal.com
We have to nail up eri and hme. ce and bge seem to autoneg fine.

(no subject)

Date: 2005-05-18 06:55 pm (UTC)
From: [identity profile] poggs.livejournal.com
I always, always, ALWAYS set things to 100Mbps full-duplex where appropriate. Gigabit appears to autonegotiate properly.

(no subject)

Date: 2005-05-19 08:55 am (UTC)
From: [identity profile] xandraius.livejournal.com
Reading that + lung butter from flu + coffee = break out the spare keyboard. Worth it though. That's it. Sun/ Cisco autoneg is a lot like a pair of very proper upper crust Englishmen... or butlers.