From 19eba6889459f4ce781861cbc5cbc9f8afa7647c Mon Sep 17 00:00:00 2001 From: Curtis Robert Date: Mon, 25 Mar 2024 08:48:00 -0700 Subject: [PATCH] [exporter/signalfx] Fix memory leaks (#31772) **Description:** Changes in PR: 1. Add correlation client Shutdown function that blocks on the waitgroup. This is the main fix of this PR that should fix the leaking goroutines. 2. Re-organize the shutdown process of the apm client correlation test suite to properly synchronize the shutting down process. 3. Fix typo 4. Only block request sender until context is cancelled. The request processor is shutdown when the context is cancelled, so this would result in `Shutdown` waiting forever, since the request would never be processed. 5. Enable goleak in some more packages. **Note**: This is contains the exact same contents as https://github.com/open-telemetry/opentelemetry-collector-contrib/pull/30887, but change number 4 is new, and should resolve the test issue the original PR was causing. **Link to tracking Issue:** Resolves https://github.com/open-telemetry/opentelemetry-collector-contrib/issues/30864 https://github.com/open-telemetry/opentelemetry-collector-contrib/issues/30438 **Testing:** All existing tests are passing, as well as added goleak checks. I'm going to run this a number of times to try to help ensure it's not flaky anymore. --- .chloggen/goleak_signalfx_correlations.yaml | 27 +++++++++++++++++++ .../internal/apm/correlations/client.go | 7 +++++ .../internal/apm/correlations/client_test.go | 23 +++++++++------- .../internal/apm/correlations/package_test.go | 14 ++++++++++ .../internal/apm/requests/sender.go | 10 +++++-- .../internal/apm/tracetracker/tracker_test.go | 3 ++- .../internal/correlation/correlation.go | 1 + .../internal/correlation/correlation_test.go | 2 ++ .../internal/correlation/package_test.go | 14 ++++++++++ 9 files changed, 88 insertions(+), 13 deletions(-) create mode 100755 .chloggen/goleak_signalfx_correlations.yaml create mode 100644 exporter/signalfxexporter/internal/apm/correlations/package_test.go create mode 100644 exporter/signalfxexporter/internal/correlation/package_test.go diff --git a/.chloggen/goleak_signalfx_correlations.yaml b/.chloggen/goleak_signalfx_correlations.yaml new file mode 100755 index 000000000000..d20b4d2ccc73 --- /dev/null +++ b/.chloggen/goleak_signalfx_correlations.yaml @@ -0,0 +1,27 @@ +# Use this changelog template to create an entry for release notes. + +# One of 'breaking', 'deprecation', 'new_component', 'enhancement', 'bug_fix' +change_type: bug_fix + +# The name of the component, or a single word describing the area of concern, (e.g. filelogreceiver) +component: signalfxexporter + +# A brief description of the change. Surround your text with quotes ("") if it needs to start with a backtick (`). +note: Fix memory leak in shutdown + +# Mandatory: One or more tracking issues related to the change. You can use the PR number here if no issue exists. +issues: [30864, 30438] + +# (Optional) One or more lines of additional information to render under the primary note. +# These lines will be padded with 2 spaces and then inserted directly into the document. +# Use pipe (|) for multiline entries. +subtext: + +# If your change doesn't affect end users or the exported elements of any package, +# you should instead start your pull request title with [chore] or use the "Skip Changelog" label. +# Optional: The change log or logs in which this entry should be included. +# e.g. '[user]' or '[user, api]' +# Include 'user' if the change is relevant to end users. +# Include 'api' if there is a change to a library API. +# Default: '[user]' +change_logs: [] diff --git a/exporter/signalfxexporter/internal/apm/correlations/client.go b/exporter/signalfxexporter/internal/apm/correlations/client.go index 427c0abb685f..ba86baf9a6c2 100644 --- a/exporter/signalfxexporter/internal/apm/correlations/client.go +++ b/exporter/signalfxexporter/internal/apm/correlations/client.go @@ -44,6 +44,7 @@ type CorrelationClient interface { Delete(*Correlation, SuccessfulDeleteCB) Get(dimName string, dimValue string, cb SuccessfulGetCB) Start() + Shutdown() } type request struct { @@ -387,3 +388,9 @@ func (cc *Client) Start() { go cc.processChan() go cc.processRetryChan() } + +// Shutdown the client. This will block until the context's cancel +// function is complete. +func (cc *Client) Shutdown() { + cc.wg.Wait() +} diff --git a/exporter/signalfxexporter/internal/apm/correlations/client_test.go b/exporter/signalfxexporter/internal/apm/correlations/client_test.go index 48cb24b996de..87ff12866cb6 100644 --- a/exporter/signalfxexporter/internal/apm/correlations/client_test.go +++ b/exporter/signalfxexporter/internal/apm/correlations/client_test.go @@ -124,7 +124,7 @@ func makeHandler(t *testing.T, corCh chan<- *request, forcedRespCode *atomic.Val }) } -func setup(t *testing.T) (CorrelationClient, chan *request, *atomic.Value, *atomic.Value, context.CancelFunc) { +func setup(t *testing.T) (CorrelationClient, *httptest.Server, chan *request, *atomic.Value, *atomic.Value, context.CancelFunc, context.Context) { serverCh := make(chan *request, 100) var forcedRespCode atomic.Value @@ -132,10 +132,6 @@ func setup(t *testing.T) (CorrelationClient, chan *request, *atomic.Value, *atom server := httptest.NewServer(makeHandler(t, serverCh, &forcedRespCode, &forcedRespPayload)) ctx, cancel := context.WithCancel(context.Background()) - go func() { - <-ctx.Done() - server.Close() - }() serverURL, err := url.Parse(server.URL) if err != nil { @@ -176,13 +172,20 @@ func setup(t *testing.T) (CorrelationClient, chan *request, *atomic.Value, *atom } client.Start() - return client, serverCh, &forcedRespCode, &forcedRespPayload, cancel + return client, server, serverCh, &forcedRespCode, &forcedRespPayload, cancel, ctx +} + +func teardown(ctx context.Context, client CorrelationClient, server *httptest.Server, serverCh chan *request, cancel context.CancelFunc) { + close(serverCh) + cancel() + <-ctx.Done() + client.Shutdown() + server.Close() } func TestCorrelationClient(t *testing.T) { - client, serverCh, forcedRespCode, forcedRespPayload, cancel := setup(t) - defer close(serverCh) - defer cancel() + client, server, serverCh, forcedRespCode, forcedRespPayload, cancel, ctx := setup(t) + defer teardown(ctx, client, server, serverCh, cancel) for _, correlationType := range []Type{Service, Environment} { for _, op := range []string{http.MethodPut, http.MethodDelete} { @@ -242,7 +245,7 @@ func TestCorrelationClient(t *testing.T) { client.Correlate(testData, CorrelateCB(func(_ *Correlation, _ error) {})) // sending the testData twice tests deduplication, since the 500 status // will trigger retries, and the requests should be deduped and the - // TotalRertriedUpdates should still only be 5 + // TotalRetriedUpdates should still only be 5 client.Correlate(testData, CorrelateCB(func(_ *Correlation, _ error) {})) cors := waitForCors(serverCh, 1, 4) diff --git a/exporter/signalfxexporter/internal/apm/correlations/package_test.go b/exporter/signalfxexporter/internal/apm/correlations/package_test.go new file mode 100644 index 000000000000..334bbbdc8b54 --- /dev/null +++ b/exporter/signalfxexporter/internal/apm/correlations/package_test.go @@ -0,0 +1,14 @@ +// Copyright The OpenTelemetry Authors +// SPDX-License-Identifier: Apache-2.0 + +package correlations + +import ( + "testing" + + "go.uber.org/goleak" +) + +func TestMain(m *testing.M) { + goleak.VerifyTestMain(m) +} diff --git a/exporter/signalfxexporter/internal/apm/requests/sender.go b/exporter/signalfxexporter/internal/apm/requests/sender.go index e06e649ea33a..708f00571655 100644 --- a/exporter/signalfxexporter/internal/apm/requests/sender.go +++ b/exporter/signalfxexporter/internal/apm/requests/sender.go @@ -48,8 +48,14 @@ func (rs *ReqSender) Send(req *http.Request) { go rs.processRequests() } - // Block until we can get through a request - rs.requests <- req + // Block until we can get the request through, or until the context is cancelled. The request processor + // shuts down when the context has been cancelled, so there's no value added to keep blocking. Blocking + // forever results in Shutdown never completing. + select { + case <-rs.ctx.Done(): + return + case rs.requests <- req: + } } } diff --git a/exporter/signalfxexporter/internal/apm/tracetracker/tracker_test.go b/exporter/signalfxexporter/internal/apm/tracetracker/tracker_test.go index 43c638ac0948..b7fb27695550 100644 --- a/exporter/signalfxexporter/internal/apm/tracetracker/tracker_test.go +++ b/exporter/signalfxexporter/internal/apm/tracetracker/tracker_test.go @@ -85,7 +85,8 @@ type correlationTestClient struct { correlateCounter int64 } -func (c *correlationTestClient) Start() { /*no-op*/ } +func (c *correlationTestClient) Start() { /*no-op*/ } +func (c *correlationTestClient) Shutdown() { /*no-op*/ } func (c *correlationTestClient) Get(_ string, dimValue string, cb correlations.SuccessfulGetCB) { atomic.AddInt64(&c.getCounter, 1) go func() { diff --git a/exporter/signalfxexporter/internal/correlation/correlation.go b/exporter/signalfxexporter/internal/correlation/correlation.go index ee00b0a470d0..281339e1df6c 100644 --- a/exporter/signalfxexporter/internal/correlation/correlation.go +++ b/exporter/signalfxexporter/internal/correlation/correlation.go @@ -137,6 +137,7 @@ func (cor *Tracker) Shutdown(_ context.Context) error { if cor != nil { if cor.correlation != nil { cor.correlation.cancel() + cor.correlation.CorrelationClient.Shutdown() } if cor.pTicker != nil { diff --git a/exporter/signalfxexporter/internal/correlation/correlation_test.go b/exporter/signalfxexporter/internal/correlation/correlation_test.go index 27633dcb2d1b..a24fb75b8b74 100644 --- a/exporter/signalfxexporter/internal/correlation/correlation_test.go +++ b/exporter/signalfxexporter/internal/correlation/correlation_test.go @@ -86,6 +86,8 @@ func TestTrackerStart(t *testing.T) { } else { require.NoError(t, err) } + + assert.NoError(t, tracker.Shutdown(context.Background())) }) } } diff --git a/exporter/signalfxexporter/internal/correlation/package_test.go b/exporter/signalfxexporter/internal/correlation/package_test.go new file mode 100644 index 000000000000..80d3ec2c9232 --- /dev/null +++ b/exporter/signalfxexporter/internal/correlation/package_test.go @@ -0,0 +1,14 @@ +// Copyright The OpenTelemetry Authors +// SPDX-License-Identifier: Apache-2.0 + +package correlation + +import ( + "testing" + + "go.uber.org/goleak" +) + +func TestMain(m *testing.M) { + goleak.VerifyTestMain(m) +}