Wait for proxy thread to terminate in integration tests#625
Merged
olix0r merged 1 commit intolinkerd:mainfrom Aug 18, 2020
Merged
Wait for proxy thread to terminate in integration tests#625olix0r merged 1 commit intolinkerd:mainfrom
olix0r merged 1 commit intolinkerd:mainfrom
Conversation
Signed-off-by: Raphael Taylor-Davies <[email protected]>
tustvold
force-pushed
the
integration-shutdown
branch
from
433d5d7 to
a9d2d93
Compare
Contributor
Author
|
CI seemed to take a rather long time, so ran some tests locally With the change, and the tests already built On main, with the tests already built So at least on my machine it doesn't seem to measurably impact the time the tests take to execute |
olix0r
approved these changes
Aug 18, 2020
Member
|
Thanks @tustvold! |
olix0r
added a commit
to linkerd/linkerd2
that referenced
this pull request
Aug 26, 2020
This release includes internal changes to the service discovery system, especially when discovering control plane components (like the destination and identity controllers). Now, the proxy attempts to balance requests across all pods in each control plane service. This requires control plane changes to use "headless" services so that SRV records are exposed. When the control plane services have a `clusterIP` set, the proxy falls back to using normal A-record lookups. --- * tracing: add richer verbose spans to http clients (linkerd/linkerd2-proxy#622) * trace: update tracing dependencies (linkerd/linkerd2-proxy#623) * Remove `Resolution` trait (linkerd/linkerd2-proxy#606) * Update proxy-identity to edge-20.8.2 (linkerd/linkerd2-proxy#627) * Add build arg for skipping identity wrapper (linkerd/linkerd2-proxy#624) * Wait for proxy thread to terminate in integration tests (linkerd/linkerd2-proxy#625) * Remove scrubbing for unused headers (linkerd/linkerd2-proxy#628) * Split orig-proto tests out of discovery tests (linkerd/linkerd2-proxy#629) * Re-enable outbound timeout test (linkerd/linkerd2-proxy#630) * profiles: perform profile resolution for IP addresses (linkerd/linkerd2-proxy#626) * Move resolve api to async-stream (linkerd/linkerd2-proxy#599) * Decouple discovery buffering from endpoint conversion (linkerd/linkerd2-proxy#631) * resolve: Add a Reset state (linkerd/linkerd2-proxy#633) * resolve: Eagerly fail resolutions (linkerd/linkerd2-proxy#634) * test: replace `net2` dependency with `socket2` (linkerd/linkerd2-proxy#635) * dns: Run DNS resolutions on the main runtime (linkerd/linkerd2-proxy#637) * Load balance requests to the control plane (linkerd/linkerd2-proxy#594) * Unify control plane client construction (linkerd/linkerd2-proxy#638)
olix0r
added a commit
to linkerd/linkerd2
that referenced
this pull request
Aug 26, 2020
This release includes internal changes to the service discovery system, especially when discovering control plane components (like the destination and identity controllers). Now, the proxy attempts to balance requests across all pods in each control plane service. This requires control plane changes to use "headless" services so that SRV records are exposed. When the control plane services have a `clusterIP` set, the proxy falls back to using normal A-record lookups. --- * tracing: add richer verbose spans to http clients (linkerd/linkerd2-proxy#622) * trace: update tracing dependencies (linkerd/linkerd2-proxy#623) * Remove `Resolution` trait (linkerd/linkerd2-proxy#606) * Update proxy-identity to edge-20.8.2 (linkerd/linkerd2-proxy#627) * Add build arg for skipping identity wrapper (linkerd/linkerd2-proxy#624) * Wait for proxy thread to terminate in integration tests (linkerd/linkerd2-proxy#625) * Remove scrubbing for unused headers (linkerd/linkerd2-proxy#628) * Split orig-proto tests out of discovery tests (linkerd/linkerd2-proxy#629) * Re-enable outbound timeout test (linkerd/linkerd2-proxy#630) * profiles: perform profile resolution for IP addresses (linkerd/linkerd2-proxy#626) * Move resolve api to async-stream (linkerd/linkerd2-proxy#599) * Decouple discovery buffering from endpoint conversion (linkerd/linkerd2-proxy#631) * resolve: Add a Reset state (linkerd/linkerd2-proxy#633) * resolve: Eagerly fail resolutions (linkerd/linkerd2-proxy#634) * test: replace `net2` dependency with `socket2` (linkerd/linkerd2-proxy#635) * dns: Run DNS resolutions on the main runtime (linkerd/linkerd2-proxy#637) * Load balance requests to the control plane (linkerd/linkerd2-proxy#594) * Unify control plane client construction (linkerd/linkerd2-proxy#638)
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Whilst implementing #601 I encountered non-determinism due to the way the integration tests don't wait for termination of the proxy thread. The changes in this PR make join_servers explicitly signal shutdown, then wait on a oneshot channel for the thread to signal draining has completed, before finally joining the thread. Simply joining the thread after signaling shutdown is insufficient as this will block the "primary" tokio runtime preventing graceful shutdown of HTTP2 connections, which rely on the clients running in the "primary" tokio runtime to be responsive.
Unfortunately the proxy appears to intentionally wait for TCP upstreams to terminate first, which requires explicit termination of the TCP client in these tests.