Score:1

NIC ports going offline just by touching the cables slightly

mx flag

A while ago we discovered an issue with most of our NICs in the data center - touching the cables just leads to the ports of the NICs going offline for a couple of seconds. All the NICs are Broadcom or Intel, all bought with servers or separately at Dell.

Last couple of months we did some tests with new servers - the first few months the NICs work without any issues, but after a couple of months, the issues become present also there. We already tried to replace cables (recommended by our provider in the data center and other people we know working as admins), but no luck - the issue persists.

Seriously, I cannot imagine, that all the NICs by popular brands become "faulty" after a couple of months, but on the other hand - trying out around 5-6 different cables by good vendors result in the same. We even bought new Broadcom NICs and installed them in one server, which was affected by this issue - brought it back online, touched the cable -> link gone for ~5 secs.

Anyone heard of this issue before or was even affected by this and found a solution?

John Mahowald avatar
cn flag
Copper or optical? Does a hardware cable tester show it in spec, or has testing been limited to NICs in servers?
mtbftw avatar
mx flag
Copper only. We also run some opticals, but no issues there. In a cable tester everything is good - we can pull/push/touch cables - they stay green. We also reported it to Dell, but they didn't hear about it before. Also our DC provider didn't.
Michael Hampton avatar
cz flag
What is the temperature on the hot side? Does it vary significantly over time?
mtbftw avatar
mx flag
The NICs are "pretty hot" under load, but I would say as usual for 10G NICs. Can't tell you the exact temperature, but I would guess around 60°C (again, under load). It doesn't vary, just staying constant. iDRACs do not report any anomaly temperatures, also we have temperature sensors in racks installed - usual rack temperature is about 21-22°C. This was our first idea: cables become unstable because of the heat. We tested it with other cables which were not installed and the issue was instantly there (without influence by the heat of the NICs).
mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.