Starting in the past 24 hours, and particularly over the last 12 hours, we have seen incrementing errors on core WMCS link between racks D5 and F4, presenting as inbound errors on cloudsw1-d5-eqiad et-0/0/53:
That port is connected to cloudsw1-f4-eqiad port et-0/0/54 over single-mode fiber. As usual with these things the likely culprit is a bad optic module at one end or other, but difficult to say which. Modules are 40G-BaseLR4 QSFP+ (blue handle I think).
DC-Ops when possible can we try replacing the module in cloudsw1-d5-eqiad et-0/0/53 and we can see if there is any improvement? The link has been drained so we can do this any time. If that does not help we can instead try swapping the far side in F4. Please ping me on irc when available and we can run some tests see how it looks. Thanks.