96 lines
5.2 KiB
YAML
96 lines
5.2 KiB
YAML
$schema: "http://json-schema.org/draft-06/schema#"
|
|
|
|
title: compute beta BackendServiceHAPolicy export schema
|
|
description: A gcloud export/import command YAML validation schema.
|
|
type: object
|
|
additionalProperties: false
|
|
properties:
|
|
COMMENT:
|
|
type: object
|
|
description: User specified info ignored by gcloud import.
|
|
additionalProperties: false
|
|
properties:
|
|
template-id:
|
|
type: string
|
|
region:
|
|
type: string
|
|
description:
|
|
type: string
|
|
date:
|
|
type: string
|
|
version:
|
|
type: string
|
|
UNKNOWN:
|
|
type: array
|
|
description: Unknown API fields that cannot be imported.
|
|
items:
|
|
type: string
|
|
fastIPMove:
|
|
description: |-
|
|
Specifies whether fast IP move is enabled, and if so, the mechanism to
|
|
achieve it. Supported values are: - DISABLED: Fast IP Move is
|
|
disabled. You can only use the haPolicy.leader API to update the
|
|
leader. - >GARP_RA: Provides a method to very quickly define a new
|
|
network endpoint as the leader. This method is faster than updating the
|
|
leader using the haPolicy.leader API. Fast IP move works as follows:
|
|
The VM hosting the network endpoint that should become the new leader
|
|
sends either a Gratuitous ARP (GARP) packet (IPv4) or an ICMPv6 Router
|
|
Advertisement(RA) packet (IPv6). Google Cloud immediately but
|
|
temporarily associates the forwarding rule IP address with that VM, and
|
|
both new and in-flight packets are quickly delivered to that VM. Note
|
|
the important properties of the Fast IP Move functionality: - The
|
|
GARP/RA-initiated re-routing stays active for approximately 20 minutes.
|
|
After triggering fast failover, you must also appropriately set the
|
|
haPolicy.leader. - The new leader instance should continue to send
|
|
GARP/RA packets periodically every 10 seconds until at least 10 minutes
|
|
after updating the haPolicy.leader (but stop immediately if it is no
|
|
longer the leader). - After triggering a fast failover, we recommend that
|
|
you wait at least 3 seconds before sending another GARP/RA packet from a
|
|
different VM instance to avoid race conditions. - Don't send GARP/RA
|
|
packets from different VM instances at the same time. If multiple
|
|
instances continue to send GARP/RA packets, traffic might be routed to
|
|
different destinations in an alternating order. This condition ceases
|
|
when a single instance issues a GARP/RA packet. - The GARP/RA
|
|
request always takes priority over the leader API. Using the
|
|
haPolicy.leader API to change the leader to a different instance will
|
|
have no effect until the GARP/RA request becomes inactive. - The
|
|
GARP/RA packets should follow the GARP/RA Packet Specifications.. -
|
|
When multiple forwarding rules refer to a regional backend service, you
|
|
need only send a GARP or RA packet for a single forwarding rule virtual
|
|
IP. The virtual IPs for all forwarding rules targeting the same backend
|
|
service will also be moved to the sender of the GARP or RA packet. The
|
|
following are the Fast IP Move limitations (that is, when fastIPMove is
|
|
not DISABLED): - Multiple forwarding rules cannot use the same IP
|
|
address if one of them refers to a regional backend service with
|
|
fastIPMove. - The regional backend service must set the network field,
|
|
and all NEGs must belong to that network. However, individual NEGs
|
|
can belong to different subnetworks of that network. - The maximum
|
|
number of network endpoints across all backends of a backend service
|
|
with fastIPMove is 32. - The maximum number of backend services with
|
|
fastIPMove that can have the same network endpoint attached to one of
|
|
its backends is 64. - The maximum number of backend services with
|
|
fastIPMove in a VPC in a region is 64. - The network endpoints that
|
|
are attached to a backend of a backend service with fastIPMove cannot
|
|
resolve to Gen3+ machines for IPv6. - Traffic directed to the leader by
|
|
a static route next hop will not be redirected to a new leader by fast
|
|
failover. Such traffic will only be redirected once an haPolicy.leader
|
|
update has taken effect. Only traffic to the forwarding rule's virtual
|
|
IP will be redirected to a new leader by fast failover.
|
|
haPolicy.fastIPMove can be set only at backend service creation time. Once
|
|
set, it cannot be updated. By default, fastIpMove is set to DISABLED.
|
|
type: string
|
|
enum:
|
|
- DISABLED
|
|
- GARP_RA
|
|
leader:
|
|
description: |-
|
|
Selects one of the network endpoints attached to the backend NEGs of this
|
|
service as the active endpoint (the leader) that receives all traffic.
|
|
When the leader changes, there is no connection draining to persist
|
|
existing connections on the old leader. You are responsible for selecting
|
|
a suitable endpoint as the leader. For example, preferring a healthy
|
|
endpoint over unhealthy ones. Note that this service does not track
|
|
backend endpoint health, and selects the configured leader
|
|
unconditionally.
|
|
$ref: BackendServiceHAPolicyLeader.yaml
|