The uplink from row D to cr1-eqiad has been disabled for some time due to an issue with VRRP. When enabling, cr1-eqiad and cr2-eqiad both become VRRP master (conflict), likely because the row D switch stack dropping VRRP packets somehow.
As we're already working on replacing the row D switch entirely, and also moving row uplinks to new line cards, we can await for that to happen - or even use the cr1-eqiad side to connect the new row switch stack first.
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | faidon | T149226 cr1-eqiad:ae4 is disabled due to VRRP issue | |||
| Resolved | • Cmjohnson | T148506 Rack and setup new eqiad row D switch stack (EX4300/QFX5100) | |||
| Unknown Object (Task) | |||||
| Unknown Object (Task) | |||||
| Unknown Object (Task) | |||||
| Unknown Object (Task) | |||||
| Resolved | None | T162681 Network maintenance on row D (databases) | |||
| Resolved | • Cmjohnson | T163002 Spread eqiad analytics Kafka nodes to multiple racks ans rows | |||
| Resolved | akosiaris | T163324 switchover icinga.wikimedia.org from einsteinium to tegmen | |||
| Resolved | akosiaris | T163326 switchover oresrdb.svc.eqiad.wmnet from oresrdb1001 to oresrdb1002 and back (after T148506) | |||
| Resolved | • Cmjohnson | T170474 Decommisson and store old row D network gear. |
This VRRP issue was the case before the cr1/2-eqiad upgrade as well, so this was likely due to some asw-d-eqiad snafu (not propagating VRRP multicasts or something.
Since the cr1-eqiad (but not the cr2-eqiad) uplinks were moved to asw2-d-eqiad last week, the VRRP issue seems to have disappeared — and with the planned replacement of asw-d-eqiad entirely, it's unlikely it will appear again.
cr1-eqiad:ae4 is now back enabled and thus row D router redundancy has been restored.