DestinationRule defines policies that apply to traffic intended for a
service after routing has occurred. These rules specify configuration
for load balancing, connection pool size from the sidecar, and outlier
detection settings to detect and evict unhealthy hosts from the load
balancing pool. For example, a simple load balancing policy for the
ratings service would look as follows:
Version specific policies can be specified by defining a named
subset and overriding the settings specified at the service level. The
following rule uses a round robin load balancing policy for all traffic
going to a subset named testversion that is composed of endpoints (e.g.,
pods) with labels (version:v3).
Note: Policies specified for subsets will not take effect until
a route rule explicitly sends traffic to this subset.
Traffic policies can be customized to specific ports as well. The
following rule uses the least connection load balancing policy for all
traffic to port 80, while uses a round robin load balancing setting for
traffic to the port 9080.
apiVersion: networking.istio.io/v1alpha3
kind: DestinationRule
metadata:name: bookinfo-ratings-port
spec:host: ratings.prod.svc.cluster.local
trafficPolicy:# Apply to all portsportLevelSettings:-port:number:80loadBalancer:simple: LEAST_CONN
-port:number:9080loadBalancer:simple: ROUND_ROBIN
apiVersion: networking.istio.io/v1beta1
kind: DestinationRule
metadata:name: bookinfo-ratings-port
spec:host: ratings.prod.svc.cluster.local
trafficPolicy:# Apply to all portsportLevelSettings:-port:number:80loadBalancer:simple: LEAST_CONN
-port:number:9080loadBalancer:simple: ROUND_ROBIN
DestinationRule
DestinationRule defines policies that apply to traffic intended for a service
after routing has occurred.
Field
Type
Description
Required
host
string
The name of a service from the service registry. Service
names are looked up from the platform’s service registry (e.g.,
Kubernetes services, Consul services, etc.) and from the hosts
declared by ServiceEntries. Rules defined for
services that do not exist in the service registry will be ignored.
Note for Kubernetes users: When short names are used (e.g. “reviews”
instead of “reviews.default.svc.cluster.local”), Istio will interpret
the short name based on the namespace of the rule, not the service. A
rule in the “default” namespace containing a host “reviews” will be
interpreted as “reviews.default.svc.cluster.local”, irrespective of
the actual namespace associated with the reviews service. To avoid
potential misconfigurations, it is recommended to always use fully
qualified domain names over short names.
Note that the host field applies to both HTTP and TCP services.
One or more named sets that represent individual versions of a
service. Traffic policies can be overridden at subset level.
No
exportTo
string[]
A list of namespaces to which this destination rule is exported.
The resolution of a destination rule to apply to a service occurs in the
context of a hierarchy of namespaces. Exporting a destination rule allows
it to be included in the resolution hierarchy for services in
other namespaces. This feature provides a mechanism for service owners
and mesh administrators to control the visibility of destination rules
across namespace boundaries.
If no namespaces are specified then the destination rule is exported to all
namespaces by default.
The value “.” is reserved and defines an export to the same namespace that
the destination rule is declared in. Similarly, the value “*” is reserved and
defines an export to all namespaces.
NOTE: in the current release, the exportTo value is restricted to
“.” or “*” (i.e., the current namespace or all namespaces).
No
TrafficPolicy
Traffic policies to apply for a specific destination, across all
destination ports. See DestinationRule for examples.
Traffic policies specific to individual ports. Note that port level
settings will override the destination-level settings. Traffic
settings specified at the destination-level will not be inherited when
overridden by port-level settings, i.e. default values will be applied
to fields omitted in port-level traffic policies.
No
Subset
A subset of endpoints of a service. Subsets can be used for scenarios
like A/B testing, or routing to a specific version of a service. Refer
to VirtualService documentation for examples of using
subsets in these scenarios. In addition, traffic policies defined at the
service-level can be overridden at a subset-level. The following rule
uses a round robin load balancing policy for all traffic going to a
subset named testversion that is composed of endpoints (e.g., pods) with
labels (version:v3).
Note: Policies specified for subsets will not take effect until
a route rule explicitly sends traffic to this subset.
One or more labels are typically required to identify the subset destination,
however, when the corresponding DestinationRule represents a host that
supports multiple SNI hosts (e.g., an egress gateway), a subset without labels
may be meaningful. In this case a traffic policy with ClientTLSSettings
can be used to identify a specific SNI host corresponding to the named subset.
Field
Type
Description
Required
name
string
Name of the subset. The service name and the subset name can
be used for traffic splitting in a route rule.
Yes
labels
map<string, string>
Labels apply a filter over the endpoints of a service in the
service registry. See route rules for examples of usage.
Traffic policies that apply to this subset. Subsets inherit the
traffic policies specified at the DestinationRule level. Settings
specified at the subset level will override the corresponding settings
specified at the DestinationRule level.
No
LoadBalancerSettings
Load balancing policies to apply for a specific destination. See Envoy’s
load balancing
documentation1
for more details.
For example, the following rule uses a round robin load balancing policy
for all traffic going to the ratings service.
The following example sets up sticky sessions for the ratings service
hashing-based load balancer for the same ratings service using the
the User cookie as the hash key.
Locality load balancer settings, this will override mesh wide settings in entirety, meaning no merging would be performed
between this object and the object one in MeshConfig
No
ConnectionPoolSettings
Connection pool settings for an upstream host. The settings apply to
each individual host in the upstream service. See Envoy’s circuit
breaker2
for more details. Connection pool settings can be applied at the TCP
level as well as at HTTP level.
For example, the following rule sets a limit of 100 connections to redis
service called myredissrv with a connect timeout of 30ms
A Circuit breaker implementation that tracks the status of each
individual host in the upstream service. Applicable to both HTTP and
TCP services. For HTTP services, hosts that continually return 5xx
errors for API calls are ejected from the pool for a pre-defined period
of time. For TCP services, connection timeouts or connection
failures to a given host counts as an error when measuring the
consecutive errors metric. See Envoy’s outlier
detection3
for more details.
The following rule sets a connection pool size of 100 HTTP1 connections
with no more than 10 req/connection to the “reviews” service. In addition,
it sets a limit of 1000 concurrent HTTP2 requests and configures upstream
hosts to be scanned every 5 mins so that any host that fails 7 consecutive
times with a 502, 503, or 504 error code will be ejected for 15 minutes.
Number of gateway errors before a host is ejected from the connection pool.
When the upstream host is accessed over HTTP, a 502, 503, or 504 return
code qualifies as a gateway error. When the upstream host is accessed over
an opaque TCP connection, connect timeouts and connection error/failure
events qualify as a gateway error.
This feature is disabled by default or when set to the value 0.
Note that consecutivegatewayerrors and consecutive5xxerrors can be
used separately or together. Because the errors counted by
consecutivegatewayerrors are also included in consecutive5xxerrors,
if the value of consecutivegatewayerrors is greater than or equal to
the value of consecutive5xxerrors, consecutivegatewayerrors will have
no effect.
Number of 5xx errors before a host is ejected from the connection pool.
When the upstream host is accessed over an opaque TCP connection, connect
timeouts, connection error/failure and request failure events qualify as a
5xx error.
This feature defaults to 5 but can be disabled by setting the value to 0.
Note that consecutivegatewayerrors and consecutive5xxerrors can be
used separately or together. Because the errors counted by
consecutivegatewayerrors are also included in consecutive5xxerrors,
if the value of consecutivegatewayerrors is greater than or equal to
the value of consecutive5xxerrors, consecutivegatewayerrors will have
no effect.
Minimum ejection duration. A host will remain ejected for a period
equal to the product of minimum ejection duration and the number of
times the host has been ejected. This technique allows the system to
automatically increase the ejection period for unhealthy upstream
servers. format: 1h/1m/1s/1ms. MUST BE >=1ms. Default is 30s.
No
maxEjectionPercent
int32
Maximum % of hosts in the load balancing pool for the upstream
service that can be ejected. Defaults to 10%.
No
minHealthPercent
int32
Outlier detection will be enabled as long as the associated load balancing
pool has at least minhealthpercent hosts in healthy mode. When the
percentage of healthy hosts in the load balancing pool drops below this
threshold, outlier detection will be disabled and the proxy will load balance
across all hosts in the pool (healthy and unhealthy). The threshold can be
disabled by setting it to 0%. The default is 0% as it’s not typically
applicable in k8s environments with few pods per service.
No
ClientTLSSettings
SSL/TLS related settings for upstream connections. See Envoy’s TLS
context5
for more details. These settings are common to both HTTP and TCP upstreams.
For example, the following rule configures a client to use mutual TLS
for connections to upstream database cluster.
Indicates whether connections to this port should be secured
using TLS. The value of this field determines how TLS is enforced.
Yes
clientCertificate
string
REQUIRED if mode is MUTUAL. The path to the file holding the
client-side TLS certificate to use.
Should be empty if mode is ISTIO_MUTUAL.
No
privateKey
string
REQUIRED if mode is MUTUAL. The path to the file holding the
client’s private key.
Should be empty if mode is ISTIO_MUTUAL.
No
caCertificates
string
OPTIONAL: The path to the file containing certificate authority
certificates to use in verifying a presented server certificate. If
omitted, the proxy will not verify the server’s certificate.
Should be empty if mode is ISTIO_MUTUAL.
No
credentialName
string
The name of the secret that holds the TLS certs for the
client including the CA certificates. Secret must exist in the
same namespace with the proxy using the certificates.
The secret (of type generic)should contain the
following keys and values: key: <privateKey>,
cert: <serverCert>, cacert: <CACertificate>.
Secret of type tls for client certificates along with
ca.crt key for CA certificates is also supported.
Only one of client certificates and CA certificate
or credentialName can be specified.
NOTE: This field is currently applicable only at gateways.
Sidecars will continue to use the certificate paths.
No
subjectAltNames
string[]
A list of alternate names to verify the subject identity in the
certificate. If specified, the proxy will verify that the server
certificate’s subject alt name matches one of the specified values.
If specified, this list overrides the value of subjectaltnames
from the ServiceEntry.
No
sni
string
SNI string to present to the server during TLS handshake.
No
LocalityLoadBalancerSetting
Locality-weighted load balancing allows administrators to control the
distribution of traffic to endpoints based on the localities of where the
traffic originates and where it will terminate. These localities are
specified using arbitrary labels that designate a hierarchy of localities in
{region}/{zone}/{sub-zone} form. For additional detail refer to
Locality Weight6
The following example shows how to setup locality weights mesh-wide.
Given a mesh with workloads and their service deployed to “us-west/zone1/”
and “us-west/zone2/”. This example specifies that when traffic accessing a
service originates from workloads in “us-west/zone1/”, 80% of the traffic
will be sent to endpoints in “us-west/zone1/”, i.e the same zone, and the
remaining 20% will go to endpoints in “us-west/zone2/”. This setup is
intended to favor routing traffic to endpoints in the same locality.
A similar setting is specified for traffic originating in “us-west/zone2/”.
If the goal of the operator is not to distribute load across zones and
regions but rather to restrict the regionality of failover to meet other
operational requirements an operator can set a ‘failover’ policy instead of
a ‘distribute’ policy.
The following example sets up a locality failover policy for regions.
Assume a service resides in zones within us-east, us-west & eu-west
this example specifies that when endpoints within us-east become unhealthy
traffic should failover to endpoints in any zone or sub-zone within eu-west
and similarly us-west should failover to us-east.
failover:-from: us-east
to: eu-west
-from: us-west
to: us-east
Optional: only one of distribute or failover can be set.
Explicitly specify loadbalancing weight across different zones and geographical locations.
Refer to Locality weighted load balancing6
If empty, the locality weight is set according to the endpoints number within it.
Optional: only failover or distribute can be set.
Explicitly specify the region traffic will land on when endpoints in local region becomes unhealthy.
Should be used together with OutlierDetection to detect unhealthy endpoints.
Note: if no OutlierDetection specified, this will not take effect.
enable locality load balancing, this is DestinationRule-level and will override mesh wide settings in entirety.
e.g. true means that turn on locality load balancing for this DestinationRule no matter what mesh wide settings is.
No
TrafficPolicy.PortTrafficPolicy
Traffic policies that apply to specific ports of the service
TLS related settings for connections to the upstream service.
No
LoadBalancerSettings.ConsistentHashLB
Consistent Hash-based load balancing can be used to provide soft
session affinity based on HTTP headers, cookies or other
properties. This load balancing policy is applicable only for HTTP
connections. The affinity to a particular destination host will be
lost when one or more hosts are added/removed from the destination
service.
The minimum number of virtual nodes to use for the hash
ring. Defaults to 1024. Larger ring sizes result in more granular
load distributions. If the number of hosts in the load balancing
pool is larger than the ring size, each host will be assigned a
single virtual node.
No
LoadBalancerSettings.ConsistentHashLB.HTTPCookie
Describes a HTTP cookie that will be used as the hash key for the
Consistent Hash load balancer. If the cookie is not present, it will
be generated.
The idle timeout for upstream connection pool connections. The idle timeout
is defined as the period in which there are no active requests.
If not set, the default is 1 hour. When the idle timeout is reached,
the connection will be closed. If the connection is an HTTP/2
connection a drain sequence will occur prior to closing the connection.
Note that request based timeouts mean that HTTP/2 PINGs will not
keep the connection alive. Applies to both HTTP1.1 and HTTP2 connections.
Specify if http1.1 connection should be upgraded to http2 for the associated destination.
No
useClientProtocol
bool
If set to true, client protocol will be preserved while initiating connection to backend.
Note that when this is set to true, h2upgradepolicy will be ineffective i.e. the client
connections will not be upgraded to http2.
No
ConnectionPoolSettings.TCPSettings.TcpKeepalive
TCP keepalive.
Field
Type
Description
Required
probes
uint32
Maximum number of keepalive probes to send without response before
deciding the connection is dead. Default is to use the OS level configuration
(unless overridden, Linux defaults to 9.)
The time duration a connection needs to be idle before keep-alive
probes start being sent. Default is to use the OS level configuration
(unless overridden, Linux defaults to 7200s (ie 2 hours.)
The time duration between keep-alive probes.
Default is to use the OS level configuration
(unless overridden, Linux defaults to 75s.)
No
LocalityLoadBalancerSetting.Distribute
Describes how traffic originating in the ‘from’ zone or sub-zone is
distributed over a set of ‘to’ zones. Syntax for specifying a zone is
{region}/{zone}/{sub-zone} and terminal wildcards are allowed on any
segment of the specification. Examples:
* - matches all localities
us-west/* - all zones and sub-zones within the us-west region
us-west/zone-1/* - all sub-zones within us-west/zone-1
Field
Type
Description
Required
from
string
Originating locality, ‘/’ separated, e.g. ‘region/zone/sub_zone’.
No
to
map<string, uint32>
Map of upstream localities to traffic distribution weights. The sum of
all weights should be 100. Any locality not present will
receive no traffic.
No
LocalityLoadBalancerSetting.Failover
Specify the traffic failover policy across regions. Since zone and sub-zone
failover is supported by default this only needs to be specified for
regions when the operator needs to constrain traffic failover so that
the default behavior of failing over to any endpoint globally does not
apply. This is useful when failing over traffic across regions would not
improve service health or may need to be restricted for other reasons
like regulatory controls.
Field
Type
Description
Required
from
string
Originating region.
No
to
string
Destination region the traffic will fail over to when endpoints in
the ‘from’ region becomes unhealthy.
No
google.protobuf.UInt32Value
Wrapper message for uint32.
The JSON representation for UInt32Value is JSON number.
Field
Type
Description
Required
value
uint32
The uint32 value.
No
LoadBalancerSettings.SimpleLB
Standard load balancing algorithms that require no tuning.
Name
Description
ROUND_ROBIN
Round Robin policy. Default
LEAST_CONN
The least request load balancer uses an O(1) algorithm which selects
two random healthy hosts and picks the host which has fewer active
requests.
RANDOM
The random load balancer selects a random healthy host. The random
load balancer generally performs better than round robin if no health
checking policy is configured.
PASSTHROUGH
This option will forward the connection to the original IP address
requested by the caller without doing any form of load
balancing. This option must be used with care. It is meant for
advanced use cases. Refer to Original Destination load balancer in
Envoy for further details.
Policy for upgrading http1.1 connections to http2.
Name
Description
DEFAULT
Use the global default.
DO_NOT_UPGRADE
Do not upgrade the connection to http2.
This opt-out option overrides the default.
UPGRADE
Upgrade the connection to http2.
This opt-in option overrides the default.
ClientTLSSettings.TLSmode
TLS connection mode
Name
Description
DISABLE
Do not setup a TLS connection to the upstream endpoint.
SIMPLE
Originate a TLS connection to the upstream endpoint.
MUTUAL
Secure connections to the upstream using mutual TLS by presenting
client certificates for authentication.
ISTIO_MUTUAL
Secure connections to the upstream using mutual TLS by presenting
client certificates for authentication.
Compared to Mutual mode, this mode uses certificates generated
automatically by Istio for mTLS authentication. When this mode is
used, all other fields in ClientTLSSettings should be empty.