User prefixes advertised via BGP to the Google Global Cache (GGC) aren't used for traditional routing. They're used to determine which prefixes should be served by the GGC node or nodes.
If the BGP session is down, the node continues to operate normally using cached prefixes.
BGP session flapped
BGP flaps are expected. The GGC BGP peer uses a virtual IP address for redundancy. During planned maintenance, this IP address may move between hosts in the node.
You don’t need to notify us about BGP flaps as long as:
- The BGP session is re-established.
- The status graph in the GGC portal shows normal traffic to users.
We recommend that you don’t alert on BGP flaps for GGC BGP sessions.
BGP session is down
We notify you if BGP to GGC is down for an extended period. Common causes of downtime are:
- BGP details in the ISP portal don't match your router configuration (IP addresses, MD5 password, etc).
- Your router doesn't have passive mode enabled.
- There's a routing issue or ACL between the GGC BGP IP and your router.