In almost a direct mirror of T374155 it seems our link from lvs1017 port enp94s0f0np0 to ssw1-e1-eqiad (see netbox) is exhibiting errors.

There is usually almost zero usage on this link, so the errors were only occasional, but fearing something similar I sent a small 10Mb stream out from the LVS, which resulted in a constant level of input errors being reported on the switch side:

Given the lack of data on this link normally I think we don't have to treat as super-urgent, but we should try to investigate as soon as possible. It does make me wonder if the SFP has also gone bad, similar to the issue with lvs1019 (maybe bad batch or just they both aged out the same?). Like the previous occasion light levels either side are good, so I less suspect an issue with the fiber run / patches but not impossible.

Either way steps would be to do a basic check on the fiber path, and if looks ok then swap one optic, then the other if it still doesn't fix. I'd say we're best tackling the lvs side first as that was what was wrong with the other one. We'll need to fail over the host before doing any of this I will work with dc-ops on irc to line up a time.