Per T363694 , we had some issues with pybal silently failing when its backends are not connected at layer 2. As pybal is unlikely to be updated, we can work around this by enabling IPIP encapsulation for our LVS pools. (See T357257 and related tickets)
Creating this ticket to:
If we decide to migrate, we'll create a separate ticket for the migration plan.
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | bking | T365616 Consider migrating Search Platform-owned clusters to IPIP encapsulation (LVS) | |||
| Resolved | bking | T367511 Migrate Cloudelastic load balancing to IPIP encapsulation (LVS) |
Per IRC conversation with @Vgutierrez , this feature is not yet available. Tagging Traffic so they can ping us when it's ready.
Now that T365689 has been completed we can discuss tackling this one @bking.
In terms of the puppet repo we have the following requirements.
We already included a sane default for profile::lvs::realserver::ipip::interfaces on common/profile/lvs/realservers/ipip.yaml, so assuming that your realservers only use $facts[interface_primary] to handle LVS traffic you won't need to override it.
Deployment would work like this:
@Vgutierrez Awesome, thank you for the comprehensive plan of action. I'll get to work on the puppet patches. Once the patches are ready, we can figure out a maintenance window if that works for you.
Change #1043302 had a related patch set uploaded (by Bking; author: Bking):
[operations/puppet@production] cloudelastic: enable IPIP for LVS