update v4 metadata to return network despite nil #2719
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
fierlion
changed the title
|
could you delete the error log on this line so this doesnt get double-logged? amazon-ecs-agent/agent/handlers/v4/container_metadata_handler.go Lines 91 to 92 in 84321fc also i would prefer to reword the error returned there^ to soemthing like |
mssrivas
approved these changes
Nov 11, 2020
angelcar
approved these changes
Nov 11, 2020
sharanyad
reviewed
Nov 11, 2020
fenxiong
pushed a commit
to fenxiong/amazon-ecs-agent
that referenced
this pull request
Nov 30, 2020
…single container (aws#2719)
sparrc
added a commit
that referenced
this pull request
Feb 5, 2021
* Update go version to 1.15 in travis.yml, github actions, dockerfiles * Retry to rename files on error as os.Rename calls are prone to failure on Windows * adding fsxwindowsfileserver task resource * Release 1.46.0 * Fix gocyclo import * fsxwindowsfileserver: fixing task concurrency issues with drive assignment * modify Makefile to use different Golang version while building Agent for Windows * Update unit test coverage minimum * update README.md to add new environment variable ECS_FSX_WINDOWS_FILE_SERVER_SUPPORTED * Explicitly initialize seelog The logger package's initialization logic is init() which will be executed by loading the package. However some of our programs are using the agent as a library and haven't configured seelog beforehand. Due to that, loading the package logs ``` 1600397429848419666 [Error] node must have children ``` which breaks the programs. I think it would be safer to do less on init() and explicitly initialize the logger from the agent's main(). Signed-off-by: Kazuyoshi Kato <[email protected]> * Release 1.47.0 * stop container timeout: increase buffer from 30s to 2m * Fix fluentbit integ test container to version 2.7.0 * improve logging for eni attachment issues * Switch back to latest firelens image in test and update the test. * save RCI client token to agent state * integ tests: cleanup and minimize dockerhub pulls (#2714) * Remove namespace tests, already converted to functional * remove taskmetadata-validator * remove GPU integ tests * remove parallel-pull-fts * remove awscli, squid * cleanup gremlin, netkitten, pause-container builds, and some unused dockerfiles * change tianon/true to gcr distroless static * remove ubuntu:20.04 and nginx:stable * remove redundant registry auth integ test * update v4 metadata to return network stats despite missing stats for single container (#2719) * removes redundant log statement (#2724) * Fix jumbled min & max for engine connection retry delays * add p4d instance type to list of GPU instances * Revert "removes redundant log statement (#2724)" This reverts commit 38ac07d. * Revert "update v4 metadata to return network stats despite missing stats for single container (#2719)" This reverts commit 7a82ee7. * Introduce an environment variable ECS_PULL_DEPENDENT_CONTAINERS_UPFRONT. The environment variable default value is true. It allows dependent containers start pulling images before dependsOn conditions in a task definition has been satisfied. Containers have known status PULLED are defined as pulled containers. These containers are stored in a pulled container map, and can be retrieved from Task Metadata Endpoint version 4 (TMDEv4) path ${ECS_CONTAINER_METADATA_URI_V4}/task when the environment variable is set to true. * load pulled containers map from state * Set ECS_PULL_DEPENDENT_CONTAINERS_UPFRONT=false as default * scripts: add ci-ecr-pull and ci-ecr-push (#2720) These scripts are for pushing the dockerhub images that we need for our ci system (integ test and build) to ECR, and for pulling those images down within the CI system. ci-ecr-pull will tag the ECR images like 'golang:1.15.4' so that users who do not run this script on their laptop or elsewhere can still pull from dockerhub, only if this script is run will the builds reference the ECR images. The reason behind this is that dockerhub is now rate-limiting anonymous pulls, so in order to avoid being rate-limited we should pull our images from ECR as much as we can. This should also speed up builds because we will be pulling our large build images (golang, gcc) from in-region s3 rather than over the internet. One side-effect of this change is that we have to reference full version numbers in our dockerfiles (1.15.4 instead of 1.15), in order to avoid querying docker for the latest version, and to avoid dockerhub pulls happening when an image is updated. * remove explicit docker pull of test registry images (#2735) Only build fluentd container on x86_64 * Update ECS_PULL_DEPENDENT_CONTAINERS_UPFRONT default value in README file * Release 1.48.0 * Add structured logging on top of seelog * Fix an edge case that causes container dependency deadlock. Scenario: two containers container A and container B, container A is a non-essential container, container B has a dependency on container A. Workflow leading to deadlock: 1. We start to pull container A's mage; 2. We fail to pull container A's image; 3. We try to create container A anyway (this is intended as we want to try cached image if possible); 4. We fail to create the container A due to missing image; 5. We try to stop container A; 6. We can't stop container A because when container B has a dependency on A, we implicitly have a reversed dependency on shutdown, such that container A won't stop before container B stops; 7. However we can't move forward container B either, because it's waiting on container A. This ends up in a deadlock. Fix: return an error when checking container B's dependency resolution when it sees that container A will never start. * Revert "save RCI client token to agent state" This reverts commit e21488b. * Revert "Add structured logging on top of seelog" This reverts commit e11ee0b. * Release 1.48.1 * fixing agent path vulnerability on windows * update v4 metadata to return network stats despite missing stats for single container (#2719) * removes redundant log statement (#2724) * Add unit test for task metadata endpoint with missing container network. * update to latest aws-sdk-go (#2752) * docker_client: Add missing error handling in getContainerStatsNotStreamed * Fetch goversion for windows from a file * Delete 'output' file (#2762) * Improve error and info logging around v2 credentials requests (#2705) also cleanup an unrelated warning message that could have previously printed the full access key id of the container instance in plain text. Signed-off-by: Cam <[email protected]> * Remove last remaining references to travis (#2763) also add static checks on 'push', so that we have test runs on the dev branch which will enable us to use the build status badge. * sync aws-sdk-go version (#2764) * Update ECS_POLL_METRICS value to be default false (#2768) * Make create container timeout configurable Co-authored-by: Mythri Garaga Manjunatha <[email protected]> Co-authored-by: Anuj Singh <[email protected]> Co-authored-by: Meghna Srivastav <[email protected]> Co-authored-by: amazon-ecs-bot <[email protected]> Co-authored-by: Kazuyoshi Kato <[email protected]> Co-authored-by: Cam <[email protected]> Co-authored-by: Sharanya Devaraj <[email protected]> Co-authored-by: Cameron Sparr <[email protected]> Co-authored-by: Ray Allan <[email protected]> Co-authored-by: Sai Prasad <[email protected]> Co-authored-by: Shubham <[email protected]> Co-authored-by: Chien-Han Lin <[email protected]> Co-authored-by: Angel Velazquez <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
This updates the v4 metadata endpoint to return network metadata for all available. Previously we threw a
http.StatusInternalServerError(500) in the case of any container in the task not having available network metadata.Implementation details
removed Error cases for individual missing metadata and added a
Warnlog level entry.Testing
make gogenerate
make test
(will update integ/functional tests as necessary given automated tests)
New tests cover the changes: no
Description for the changelog
Update v4 metadata to return network despite nil
Licensing
By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.