release: flipping deprecated features to be fatal-by-default by alyssawilk · Pull Request #7549 · envoyproxy/envoy

alyssawilk · 2019-07-11T20:26:50Z

Flipping the following deprecated fields to be fatal-by-default

runtime from bootstrap.proto
catch_all_cluster from redis_proxy.proto
use_original_dst from lds.proto
max_stats from server_info.proto
cluster from redis_proxy.proto
max_obj_name_len from server_info.proto

Fixing a typo in release instructions
Undeprecating hosts as we're not going to remove it until v3.
Fixing canonical configs to use legal fields (see #7548 for forward-fixing this)

Risk Level: High (for folks ignoring their deprecation warnings)
Testing: //test/... (fixed canonical configs)
Docs Changes: no (should there be?)
Release Notes: no (should there be?)

Signed-off-by: Alyssa Wilk <alyssar@chromium.org>

alyssawilk · 2019-07-11T20:28:21Z

@htuch for API stuff and layered runtime
@envoyproxy/senior-maintainers for any objection to fatal-by-defaults

This PR should land with an envoy-announce email so please don't merge it unless you want to send that email :-)
Given it's pretty close to EoD Thursday I'll likely merge it Monday.

mattklein123

Thanks for staying on top of this. Few small comments.

/wait

mattklein123 · 2019-07-11T20:51:28Z

source/common/runtime/runtime_features.cc

+    "envoy.deprecated_features.bootstrap.proto:runtime",
+    "envoy.deprecated_features.redis_proxy.proto:catch_all_cluster",
+    "envoy.deprecated_features.lds.proto:use_original_dst",
+    "envoy.deprecated_features.server_info.proto:max_stats",


I think this and below are on output protos so they won't have any effect. Not sure how we could know this. Thoughts?

Good point. No ideas here - I'd be inclined to leave them in to avoid script churn.
Maybe have file or proto naming in v3 to make things more clear going forward?
@htuch

mattklein123 · 2019-07-11T20:52:34Z

configs/envoy_double_proxy_v2.template.yaml

-  symlink_root: "/srv/runtime_data/current"
-  subdirectory: envoy
-  override_subdirectory: envoy_override
+layered_runtime:


The equivalent of what we had before would be:

layered_runtime: layers: - name: root disk_layer: symlink_root: /srv/configset/envoydata/current subdirectory: envoy - name: override disk_layer: symlink_root: /srv/configset/envoydata/current subdirectory: envoy_override append_service_cluster: true - name: admin admin_layer: {}

Should we keep it equivalent?

mattklein123 · 2019-07-11T20:52:49Z

configs/envoy_front_proxy_v2.template.yaml

-  subdirectory: envoy
-  override_subdirectory: envoy_override
+layered_runtime:
+    layers:


see above comment, also indent?

mattklein123 · 2019-07-11T20:52:58Z

configs/envoy_service_to_service_v2.template.yaml

-  subdirectory: envoy
-  override_subdirectory: envoy_override
+layered_runtime:
+    layers:


mattklein123 · 2019-07-15T16:25:51Z

source/common/runtime/runtime_features.cc

+    "envoy.deprecated_features.lds.proto:use_original_dst",
+    "envoy.deprecated_features.server_info.proto:max_stats",
+    "envoy.deprecated_features.redis_proxy.proto:cluster",
+    "envoy.deprecated_features.server_info.proto:max_obj_name_len",


I think we need

envoy/source/common/config/utility.cc

Line 302 in b3f848d

"envoy.deprecated_features.v1_filter_json_config")) {

also?

Oh man, I totally missed this in review but this should have been envoy.reloadable_features.v1_filter_json_config which is why it wasn't caught.

I tried flipping it, and it broke the bootstrap test pretty hard. Beyond needing the same runtime layers update, there's some issue with proto constraint which I've tracked down to a really confusing red herring where when we to do proto to json conversion we end up failing and so having a completely empty proto, so failing the first validator which checks for non-empty bytes

Given this didn't land as a config change what do we think of renaming, and flipping in its own PR, ideally after some digging into the hot restart test?

Sure whatever you think is best. I mainly want to get this deprecation moving since this will allow us to delete a ton of code (all of v1).

Would you have cycles to look into what's going on with the filter config? I was going to ask we assign whoever introduced the variable to sorting out the test and looks like it was you :-P

Yes, sorry, what do you want me to do exactly? If you give me a diff that is broken I can debug it. :)

I think we land this as-is, and you flip the v1 json separately, since it currently doesn't pass tests?

OK got it, but you want me to change the name also?

Signed-off-by: Alyssa Wilk <alyssar@chromium.org>

mattklein123

LGTM!

mattklein123 · 2019-07-15T21:27:45Z

source/common/runtime/runtime_features.cc

    // Acts as both a test entry for deprecated.proto and a marker for the Envoy
    // deprecation scripts.
    "envoy.deprecated_features.deprecated.proto:is_deprecated_fatal",
+    "envoy.deprecated_features.bootstrap.proto:runtime",


nit: I would probably alpha sort all of these if possible.

I will alpha order when I do my follow up PR.

"use_original_dst" and "bind_to_port" are two complementary parts of the "virtual listeners" feature, and they should be deprecated together. However, they are both currently exempted (see: envoyproxy#5355), so revert the change from envoyproxy#7549. Signed-off-by: Piotr Sikora <piotrsikora@google.com>

"use_original_dst" and "bind_to_port" are two complementary parts of the "virtual listeners" feature, and they should be deprecated together. However, they are both currently exempted (see: #5355), so revert the change from #7549. Signed-off-by: Piotr Sikora <piotrsikora@google.com>

"use_original_dst" and "bind_to_port" are two complementary parts of the "virtual listeners" feature, and they should be deprecated together. However, they are both currently exempted (see: envoyproxy#5355), so revert the change from envoyproxy#7549. Signed-off-by: Piotr Sikora <piotrsikora@google.com>

release: flipping deprecated features to be fatal-by-default

e988d1d

Signed-off-by: Alyssa Wilk <alyssar@chromium.org>

alyssawilk assigned htuch Jul 11, 2019

mattklein123 requested changes Jul 11, 2019

View reviewed changes

repokitteh-read-only bot added the waiting label Jul 11, 2019

mattklein123 reviewed Jul 15, 2019

View reviewed changes

canonical config

8d942c9

Signed-off-by: Alyssa Wilk <alyssar@chromium.org>

repokitteh-read-only bot removed the waiting label Jul 15, 2019

mattklein123 approved these changes Jul 15, 2019

View reviewed changes

mattklein123 merged commit 7a55417 into envoyproxy:master Jul 16, 2019

PiotrSikora mentioned this pull request Jul 17, 2019

release: un-deprecate "use_original_dst". #7623

Merged

alyssawilk deleted the release3 branch July 31, 2019 20:33

Conversation

alyssawilk commented Jul 11, 2019

Uh oh!

alyssawilk commented Jul 11, 2019

Uh oh!

mattklein123 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mattklein123 left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants