1.3.1 2017-10-24

----------------

* Kubernetes
  * Fixed a failure to update routing data after restarting watches (linkerd#1674).
  * Ensured that Kubernetes API watch events earlier than the current state are ignored (linkerd#1681).
* Added support for Istio Mixer precondition checks (linkerd#1606).
* Removed spurious error message logging from Consul namer (linkerd#1682).
* Changed DNS SRV record namer to use system DNS resolver configuration (linkerd#1679).
* Added `timestampHeader` configuration to support New Relic request queue (linkerd#1672).

1.3.1-rc1

=========

Release candidate with potential fix for issues affecting Kubernetes watches.

1.3.0 (linkerd#1663)

## 1.3.0 2017-10-06

* **Breaking Change**: All HTTP engines are now Netty 4; `engine:` configuration key is no longer valid.
* Upgraded to Finagle 7.1
* Kubernetes
  * Added a workaround for an issue where Kubernetes namers fail to update because watches are not correctly restarted due to a regression in some versions of Kubernetes (linkerd#1636).
  * Fixed `io.l5d.k8s.configMap` interpreter failing to update after receiving an invalid dtab (linkerd#1639).
  * Performance improvements for Kubernetes namers.
* Prometheus
  * Added an optional `prefix:` configuration key to add a prefix to all metrics reported by Linkerd (linkerd#1655).
* DNS SRV Record namer
  * Ensured that DNS names in SRV queries are absolute (linkerd#1637).
  * Added an optional `domain` config key for relative DNS lookups (linkerd#1637).
  * Removed redundant `dnssrv` metrics scope from SRV record namer metrics (linkerd#1637).
* Consul
  * Consul namers no longer watch the entire list of services, improving performance significantly when there are large numbers of services (linkerd#1646).
* Curator
  * Added support for `ServiceInstance` objects with custom payloads (linkerd#1272).

1.2.1 2017-09-14

----------------

Fix for an issue where Kubernetes namers would continue to route to old
endpoints after a service was deleted and re-created, or scaled down to 0 and
then scaled back up.

Also includes:
* The path on which the Prometheus telemeter serves metrics can now be
  set in the config file.
* Minor documentation fixes.

1.2.0 2017-09-07

----------------

* **Breaking Change**: `io.l5d.mesh`, `io.l5d.thriftNameInterpreter`, linkerd
  admin, and namerd admin now serve on 127.0.0.1 by default (instead of
  0.0.0.0).
* **Breaking Change**: Removed support for PKCS#1-formatted keys. PKCS#1 formatted keys must be converted to PKCS#8 format.
* Added experimental `io.l5d.dnssrv` namer for DNS SRV records
* Kubernetes
  * Added an experimental `io.l5d.k8s.configMap` interpreter for reading dtabs from a Kubernetes ConfigMap . This interpreter will respond to changes in the ConfigMap, allowing for dynamic dtab updates without the need to run Namerd.
  * Made ingress controller's ingress class annotation configurable .
  * Fixed an issue where Linkerd would continue routing traffic to endpoints of a service after that service was removed .
  * Major refactoring and performance improvements to `io.l5d.k8s` and `io.l5d.k8s.ns` namers .
  * Ingress controller now checks all available ingress resources before using a default backend .
  * Ingress controller now correctly routes requests with host headers that contain ports .
* HTTP/2
  * Fixed an issue where long-running H2 streams would eventually hang .
  * Fixed a memory leak on long-running H2 streams
  * Added a user-friendly error message when a HTTP/2 router receives a HTTP/1 request
* HTTP/1
  * Removed spurious `ReaderDiscarded` exception logged on HTTP/1 retries
* Consul
  * Added support for querying Consul by specific service health states
  * Consul namers and Dtab store now fall back to a last known good state on Consul observation errors
  * Improved log messages for Consul observation errors
* TLS
  * Removed support for PKCS#1 keys
  * Added validation to prevent incompatible `disableValidation: true` and `clientAuth` settings in TLS client configurations
* Changed `io.l5d.mesh`, `io.l5d.thriftNameInterpreter`, linkerd
  admin, and namerd admin to serve on 127.0.0.1 by default (instead of
  0.0.0.0)
* Deprecated `io.l5d.statsd` telemeter.

1.2.0-rc2

---------

+ Kubernetes
  - Update default backend behavior and strip host ports (linkerd#1607)
+ H2
  - Ensure frames are released after classification timeout (linkerd#1598)
  - Fix memory leak in long-running H2 streams (linkerd#1613)

1.2.0-rc1

Release candidate for testing Kubernetes changes.

1.1.3 (linkerd#1574)

The 1.1.3 release of Linkerd is mostly focused on improving our HTTP/2 support,
including better support for gRPC.  Linkerd now supports automatic retries in
HTTP/2 for retryable requests.

* HTTP/2
  * Cleaned up spurious errors messages in the Linkerd log output.
  * Added a number of gRPC response classifiers that use the `grpc-status` code
    to determine if the response was successful and if it should be retried.
    See [the docs](https://linkerd.io/config/1.1.3/linkerd/index.html#grpc-response-classifiers) for details.
  * Added support for failure accrual and automatic retries to HTTP/2.
  * Fixed a memory leak related to messages with only a headers frame.
* Istio
  * Added HTTP/2 support to the Istio integration: the `io.l5d.k8s.istio`
    identifier can now be used in H2 router configs.
  * Added support for HTTPRedirect Route Rules.
* The Linkerd and Namerd admin sites can now be configured to require HTTPS.

1.1.2 (linkerd#1499) (linkerd#1499)

* Marathon Namer TLS support, for DC/OS strict mode.
* We fixed an issue where requests that time out were not being retried.
* HTTP 1.1 protocol fixes for chunked transfer encoding and
* `Content-Length`.
* Improved memory allocation in InfluxDb and Prometheus telemeters.
* Documentation fixes.

1.1.1 (linkerd#1485)

This is a big release with lots of fun stuff inside.

We've added some new features!
* Linkerd now features integration with Istio! (Beta.) This is a big feature. Blog post coming soon.
* We've introduced a new request logger plugin interface, for plugins that take an action (such as logging) on each request. This is currently used by the Istio plugin to report metadata about each request.

We’ve fixed some things!
* We fixed a connection leak in HTTP/2 by properly multiplexing streams over a single connection.
* The configured failure accrual backoff parameter was being ignored. Now it's not!
* We fixed a TLS issue when no trust certs were specified.  As a result, using TLS with egress now works again.
* We fixed an exception when a Kubernetes Service's `targetPort` value is returned as a name instead of a number.
* The admin dashboard now displays server connections, standardizing client and server displays.

We’ve made some internal changes to keep up with the latest and greatest:
* Netty4 is now the default engine for HTTP.
* We’ve upgrade to Finagle 6.45 under the hood.