Subsetting(mapping=None, *, ignore_unknown_fields=False, **kwargs)Subsetting configuration for this BackendService. Currently this is applicable only for Internal TCP/UDP load balancing, Internal HTTP(S) load balancing and Traffic Director.
.. _oneof: https://proto-plus-python.readthedocs.io/en/stable/fields.html#oneofs-mutually-exclusive-fields
Attributes |
|
|---|---|
| Name | Description |
policy |
str
Check the Policy enum for the list of possible values. This field is a member of oneof_ _policy.
|
subset_size |
int
The number of backends per backend group assigned to each proxy instance or each service mesh client. An input parameter to the CONSISTENT_HASH_SUBSETTING
algorithm. Can only be set if policy is set to
CONSISTENT_HASH_SUBSETTING. Can only be set if load
balancing scheme is INTERNAL_MANAGED or
INTERNAL_SELF_MANAGED.
subset_size is optional for Internal HTTP(S) load
balancing and required for Traffic Director.
If you do not provide this value, Cloud Load Balancing will
calculate it dynamically to optimize the number of
proxies/clients visible to each backend and vice versa.
Must be greater than 0. If subset_size is larger than
the number of backends/endpoints, then subsetting is
disabled.
This field is a member of oneof_ _subset_size.
|
Classes
Policy
Policy(value) For Traffic Director, the number of backends per backend
group (the subset size) is based on the `subset_size`
parameter.
For Internal HTTP(S) load balancing, the number of backends
per backend group (the subset size) is dynamically adjusted
in two cases:
- As the number of proxy instances participating in Internal
HTTP(S) load balancing increases, the subset size
decreases.
- When the total number of backends in a network exceeds the
capacity of a single proxy instance, subset sizes are
reduced automatically for each service that has backend
subsetting enabled.
NONE (2402104):
No Subsetting.
Clients may open connections and send traffic to
all backends of this backend service. This can
lead to performance issues if there is
substantial imbalance in the count of clients
and backends.