Ensure proper gofail package version in robustness tests #18397

serathius · 2024-08-02T07:25:05Z

Addition of tools/mod package into older branches moved the source of truth for gofail version, so the existing makefile code stopped overriding changes. This resulted in robustness tests flaking with release of [email protected] that allowed new type of failpoint.

This PR should help us ensure that gofail version is set properly on older branches.

Before we see downgraded go.etcd.io/gofail v0.2.0 => v0.1.0 line in the following build logs

$ make /tmp/etcd-release-3.5-failpoints/bin 
rm -rf /tmp/etcd-release-3.5-failpoints/
mkdir -p /tmp/etcd-release-3.5-failpoints/
cd /tmp/etcd-release-3.5-failpoints/; \
  git clone --depth 1 --branch release-3.5 https://github.com/etcd-io/etcd.git .; \
  go get go.etcd.io/[email protected]; \
  (cd server; go get go.etcd.io/[email protected]); \
  (cd etcdctl; go get go.etcd.io/[email protected]); \
  (cd etcdutl; go get go.etcd.io/[email protected]); \
  FAILPOINTS=true ./build;
Cloning into '.'...
remote: Enumerating objects: 1656, done.
remote: Counting objects: 100% (1656/1656), done.
remote: Compressing objects: 100% (1465/1465), done.
remote: Total 1656 (delta 355), reused 667 (delta 149), pack-reused 0
Receiving objects: 100% (1656/1656), 4.26 MiB | 15.26 MiB/s, done.
Resolving deltas: 100% (355/355), done.
go: upgraded go.etcd.io/gofail v0.1.0 => v0.2.0
go: upgraded github.com/stretchr/testify v1.8.4 => v1.9.0
go: upgraded go.etcd.io/gofail v0.1.0 => v0.2.0
DEPRECATED!!! Use build.sh script instead.

% 'gofail' 'enable' 'server/etcdserver/' 'server/lease/leasehttp' 'server/mvcc/' 'server/wal/' 'server/mvcc/backend/'
go: removed go.etcd.io/etcd/etcdctl/v3 v3.5.15
go: removed go.etcd.io/etcd/etcdutl/v3 v3.5.15
go: removed go.etcd.io/etcd/server/v3 v3.5.15
go: removed go.etcd.io/etcd/tests/v3 v3.5.15
go: downgraded go.etcd.io/gofail v0.2.0 => v0.1.0
go: downgraded go.etcd.io/gofail v0.2.0 => v0.1.0
go: downgraded go.etcd.io/gofail v0.2.0 => v0.1.0
go: downgraded go.etcd.io/gofail v0.2.0 => v0.1.0
go: upgraded github.com/stretchr/testify v1.8.4 => v1.9.0
% 'gofail' 'enable' 'server/etcdserver/' 'server/lease/leasehttp' 'server/mvcc/' 'server/wal/' 'server/mvcc/backend/'
% 'rm' '-f' 'bin/etcd'
% (cd server && 'env' 'CGO_ENABLED=0' 'GO_BUILD_FLAGS=' 'GOOS=linux' 'GOARCH=amd64' 'go' 'build' '-trimpath' '-installsuffix=cgo' '-ldflags=-X=go.etcd.io/etcd/api/v3/version.GitSHA=6abcc18-FAILPOINTS' '-o=../bin/etcd' '.')
% 'rm' '-f' 'bin/etcdutl'
% (cd etcdutl && 'env' 'GO_BUILD_FLAGS=' 'CGO_ENABLED=0' 'GO_BUILD_FLAGS=' 'GOOS=linux' 'GOARCH=amd64' 'go' 'build' '-trimpath' '-installsuffix=cgo' '-ldflags=-X=go.etcd.io/etcd/api/v3/version.GitSHA=6abcc18-FAILPOINTS' '-o=../bin/etcdutl' '.')
% 'rm' '-f' 'bin/etcdctl'
% (cd etcdctl && 'env' 'GO_BUILD_FLAGS=' 'CGO_ENABLED=0' 'GO_BUILD_FLAGS=' 'GOOS=linux' 'GOARCH=amd64' 'go' 'build' '-trimpath' '-installsuffix=cgo' '-ldflags=-X=go.etcd.io/etcd/api/v3/version.GitSHA=6abcc18-FAILPOINTS' '-o=../bin/etcdctl' '.')
SUCCESS: etcd_build (GOARCH=amd64)

With the change we no longer get it:

make /tmp/etcd-release-3.5-failpoints/bin 
rm -rf /tmp/etcd-release-3.5-failpoints/
mkdir -p /tmp/etcd-release-3.5-failpoints/
cd /tmp/etcd-release-3.5-failpoints/; \
  git clone --depth 1 --branch release-3.5 https://github.com/etcd-io/etcd.git .; \
  go get go.etcd.io/[email protected]; \
  (cd tools/mod; go get go.etcd.io/[email protected]); \
  FAILPOINTS=true ./build;
Cloning into '.'...
remote: Enumerating objects: 1656, done.
remote: Counting objects: 100% (1656/1656), done.
remote: Compressing objects: 100% (1465/1465), done.
remote: Total 1656 (delta 355), reused 667 (delta 149), pack-reused 0
Receiving objects: 100% (1656/1656), 4.26 MiB | 9.02 MiB/s, done.
Resolving deltas: 100% (355/355), done.
go: upgraded go.etcd.io/gofail v0.1.0 => v0.2.0
go: upgraded go.etcd.io/gofail v0.1.0 => v0.2.0
DEPRECATED!!! Use build.sh script instead.

% 'gofail' 'enable' 'server/etcdserver/' 'server/lease/leasehttp' 'server/mvcc/' 'server/wal/' 'server/mvcc/backend/'
go: upgraded github.com/stretchr/testify v1.8.4 => v1.9.0
go: upgraded go.etcd.io/gofail v0.1.0 => v0.2.0
go: upgraded github.com/stretchr/testify v1.8.4 => v1.9.0
go: upgraded go.etcd.io/gofail v0.1.0 => v0.2.0
% 'gofail' 'enable' 'server/etcdserver/' 'server/lease/leasehttp' 'server/mvcc/' 'server/wal/' 'server/mvcc/backend/'
% 'rm' '-f' 'bin/etcd'
% (cd server && 'env' 'CGO_ENABLED=0' 'GO_BUILD_FLAGS=' 'GOOS=linux' 'GOARCH=amd64' 'go' 'build' '-trimpath' '-installsuffix=cgo' '-ldflags=-X=go.etcd.io/etcd/api/v3/version.GitSHA=6abcc18-FAILPOINTS' '-o=../bin/etcd' '.')
% 'rm' '-f' 'bin/etcdutl'
% (cd etcdutl && 'env' 'GO_BUILD_FLAGS=' 'CGO_ENABLED=0' 'GO_BUILD_FLAGS=' 'GOOS=linux' 'GOARCH=amd64' 'go' 'build' '-trimpath' '-installsuffix=cgo' '-ldflags=-X=go.etcd.io/etcd/api/v3/version.GitSHA=6abcc18-FAILPOINTS' '-o=../bin/etcdutl' '.')
% 'rm' '-f' 'bin/etcdctl'
% (cd etcdctl && 'env' 'GO_BUILD_FLAGS=' 'CGO_ENABLED=0' 'GO_BUILD_FLAGS=' 'GOOS=linux' 'GOARCH=amd64' 'go' 'build' '-trimpath' '-installsuffix=cgo' '-ldflags=-X=go.etcd.io/etcd/api/v3/version.GitSHA=6abcc18-FAILPOINTS' '-o=../bin/etcdctl' '.')
SUCCESS: etcd_build (GOARCH=amd64)

Signed-off-by: Marek Siarkowicz <[email protected]>

serathius · 2024-08-02T07:25:20Z

cc @ah8ad3 @ahrtr @siyuanfoundation @fuweid @MadhavJivrajani

henrybear327 · 2024-08-02T07:57:50Z

tests/robustness/makefile.mk

@@ -124,9 +125,7 @@ $(GOPATH)/bin/gofail: tools/mod/go.mod tools/mod/go.sum
 	cd /tmp/etcd-release-3.5-failpoints/; \
 	  git clone --depth 1 --branch release-3.5 https://github.com/etcd-io/etcd.git .; \
 	  go get go.etcd.io/gofail@${GOFAIL_VERSION}; \
-	  (cd server; go get go.etcd.io/gofail@${GOFAIL_VERSION}); \


Is there a reason that for /tmp/etcd-v3.5.12-beforeSendWatchResponse/bin we keep

(cd server; go get go.etcd.io/gofail@${GOFAIL_VERSION}); \ (cd etcdctl; go get go.etcd.io/gofail@${GOFAIL_VERSION}); \ (cd etcdutl; go get go.etcd.io/gofail@${GOFAIL_VERSION}); \

but we remove those for /tmp/etcd-release-3.5-failpoints/bin? :)

We know that release branch has the tools/mod directory, however for tags we don't know. Some of them might have it, some don't.

But we will always have server, etcdctl, and etcdutl directories, so we will always need to go get the gofail library for them, no? :)

The build.sh will execute the go get go.etcd.io/gofail@${GOFAIL_VERSION}) commands for all servers, etcdutl, etcdctl and tests., so it's OK to remove the duplicated commands in makefile.

etcd/build.sh

Lines 33 to 36 in d83c8bd

cd ./server && go get go.etcd.io/gofail@"${GOFAIL_VERSION}"

cd ../etcdutl && go get go.etcd.io/gofail@"${GOFAIL_VERSION}"

cd ../etcdctl && go get go.etcd.io/gofail@"${GOFAIL_VERSION}"

cd ../tests && go get go.etcd.io/gofail@"${GOFAIL_VERSION}"

ahrtr · 2024-08-02T08:53:10Z

Before we see downgraded go.etcd.io/gofail v0.2.0 => v0.1.0 line in the following build logs

I did not reproduce this. The following is what I saw in my local environment,

$ make /tmp/etcd-release-3.5-failpoints/bin
go install go.etcd.io/[email protected]
rm -rf /tmp/etcd-release-3.5-failpoints/
mkdir -p /tmp/etcd-release-3.5-failpoints/
cd /tmp/etcd-release-3.5-failpoints/; \
	  git clone --depth 1 --branch release-3.5 https://github.com/etcd-io/etcd.git .; \
	  go get go.etcd.io/[email protected]; \
	  (cd server; go get go.etcd.io/[email protected]); \
	  (cd etcdctl; go get go.etcd.io/[email protected]); \
	  (cd etcdutl; go get go.etcd.io/[email protected]); \
	  FAILPOINTS=true ./build;
Cloning into '.'...
remote: Enumerating objects: 1656, done.
remote: Counting objects: 100% (1656/1656), done.
remote: Compressing objects: 100% (1473/1473), done.
remote: Total 1656 (delta 356), reused 654 (delta 141), pack-reused 0
Receiving objects: 100% (1656/1656), 4.26 MiB | 4.58 MiB/s, done.
Resolving deltas: 100% (356/356), done.
go: upgraded go.etcd.io/gofail v0.1.0 => v0.2.0
\e[91mDEPRECATED!!! Use build.sh script instead.\e[0m

% gofail enable server/etcdserver/ server/lease/leasehttp server/mvcc/ server/wal/ server/mvcc/backend/
% gofail enable server/etcdserver/ server/lease/leasehttp server/mvcc/ server/wal/ server/mvcc/backend/
% rm -f bin/etcd
% (cd server && env CGO_ENABLED=0 GO_BUILD_FLAGS= GOOS=darwin GOARCH=arm64 go build -trimpath -installsuffix=cgo -ldflags=-X=go.etcd.io/etcd/api/v3/version.GitSHA=d83c8bd-FAILPOINTS -o=../bin/etcd .)
% rm -f bin/etcdutl
% (cd etcdutl && env GO_BUILD_FLAGS= CGO_ENABLED=0 GO_BUILD_FLAGS= GOOS=darwin GOARCH=arm64 go build -trimpath -installsuffix=cgo -ldflags=-X=go.etcd.io/etcd/api/v3/version.GitSHA=d83c8bd-FAILPOINTS -o=../bin/etcdutl .)
% rm -f bin/etcdctl
% (cd etcdctl && env GO_BUILD_FLAGS= CGO_ENABLED=0 GO_BUILD_FLAGS= GOOS=darwin GOARCH=arm64 go build -trimpath -installsuffix=cgo -ldflags=-X=go.etcd.io/etcd/api/v3/version.GitSHA=d83c8bd-FAILPOINTS -o=../bin/etcdctl .)
SUCCESS: etcd_build (GOARCH=arm64)

The only reason I can think of is that the GOFAIL_VERSION was v0.1.0 when executing the command go get go.etcd.io/gofail@${GOFAIL_VERSION}.

serathius · 2024-08-02T08:56:40Z

Hmm, worrying inconsistency. The problem happens in my environment and in CI like in Job https://prow.k8s.io/view/gs/kubernetes-jenkins/logs/ci-etcd-robustness-main-amd64/1819076252212924416, for which you can find logs in https://storage.googleapis.com/kubernetes-jenkins/logs/ci-etcd-robustness-main-amd64/1819076252212924416/build-log.txt

ahrtr · 2024-08-02T09:14:23Z

The problem happens in my environment

Please execute command below and let me know what you get,

$ cd /tmp/etcd-release-3.5-failpoints
$ go list -m -f '{{.Version}}' go.etcd.io/gofail
v0.2.0

cleanup /tmp/etcd-release-3.5-failpoints and try again?

ahrtr · 2024-08-02T09:16:28Z

Sorry, please execute commands below instead,

$ cd /tmp/etcd-release-3.5-failpoints
$ cd tools/mod && go list -m -f '{{.Version}}' go.etcd.io/gofail
v0.2.0

serathius · 2024-08-02T09:33:23Z

Oh, with #18395 merged you cannot test it on branch target like /tmp/etcd-release-3.5-failpoints/bin. You can still do that on /tmp/etcd-v3.5.15-failpoints/bin target.

See my results:

etcd $ git checkout origin/main
HEAD is now at 8ac9ce832 Merge pull request #18392 from serathius/robustness-deflake-watch-progress
etcd $ rm -rf /tmp/etcd-v3.5.15-failpoints/bin/
etcd $ make /tmp/etcd-v3.5.15-failpoints/bin
rm -rf /tmp/etcd-v3.5.15-failpoints/
mkdir -p /tmp/etcd-v3.5.15-failpoints/
cd /tmp/etcd-v3.5.15-failpoints/; \
  git clone --depth 1 --branch v3.5.15 https://github.com/etcd-io/etcd.git .; \
  go get go.etcd.io/[email protected]; \
  (cd server; go get go.etcd.io/[email protected]); \
  (cd etcdctl; go get go.etcd.io/[email protected]); \
  (cd etcdutl; go get go.etcd.io/[email protected]); \
  FAILPOINTS=true ./build;
Cloning into '.'...
remote: Enumerating objects: 1664, done.
remote: Counting objects: 100% (1664/1664), done.
remote: Compressing objects: 100% (1472/1472), done.
remote: Total 1664 (delta 357), reused 672 (delta 150), pack-reused 0
Receiving objects: 100% (1664/1664), 4.26 MiB | 9.17 MiB/s, done.
Resolving deltas: 100% (357/357), done.
Note: switching to '9a5533382d84999e4e79642e1ec0f8bfa9b70ba8'.

You are in 'detached HEAD' state. You can look around, make experimental
changes and commit them, and you can discard any commits you make in this
state without impacting any branches by switching back to a branch.

If you want to create a new branch to retain commits you create, you may
do so (now or later) by using -c with the switch command. Example:

  git switch -c <new-branch-name>

Or undo this operation with:

  git switch -

Turn off this advice by setting config variable advice.detachedHead to false

go: upgraded go.etcd.io/gofail v0.1.0 => v0.2.0
go: upgraded github.com/stretchr/testify v1.8.4 => v1.9.0
go: upgraded go.etcd.io/gofail v0.1.0 => v0.2.0
DEPRECATED!!! Use build.sh script instead.

% 'gofail' 'enable' 'server/etcdserver/' 'server/lease/leasehttp' 'server/mvcc/' 'server/wal/' 'server/mvcc/backend/'
go: removed go.etcd.io/etcd/etcdctl/v3 v3.5.15
go: removed go.etcd.io/etcd/etcdutl/v3 v3.5.15
go: removed go.etcd.io/etcd/server/v3 v3.5.15
go: removed go.etcd.io/etcd/tests/v3 v3.5.15
go: downgraded go.etcd.io/gofail v0.2.0 => v0.1.0
go: downgraded go.etcd.io/gofail v0.2.0 => v0.1.0
go: downgraded go.etcd.io/gofail v0.2.0 => v0.1.0
go: downgraded go.etcd.io/gofail v0.2.0 => v0.1.0
go: upgraded github.com/stretchr/testify v1.8.4 => v1.9.0
% 'gofail' 'enable' 'server/etcdserver/' 'server/lease/leasehttp' 'server/mvcc/' 'server/wal/' 'server/mvcc/backend/'
% 'rm' '-f' 'bin/etcd'
% (cd server && 'env' 'CGO_ENABLED=0' 'GO_BUILD_FLAGS=' 'GOOS=linux' 'GOARCH=amd64' 'go' 'build' '-trimpath' '-installsuffix=cgo' '-ldflags=-X=go.etcd.io/etcd/api/v3/version.GitSHA=9a55333-FAILPOINTS' '-o=../bin/etcd' '.')
% 'rm' '-f' 'bin/etcdutl'
% (cd etcdutl && 'env' 'GO_BUILD_FLAGS=' 'CGO_ENABLED=0' 'GO_BUILD_FLAGS=' 'GOOS=linux' 'GOARCH=amd64' 'go' 'build' '-trimpath' '-installsuffix=cgo' '-ldflags=-X=go.etcd.io/etcd/api/v3/version.GitSHA=9a55333-FAILPOINTS' '-o=../bin/etcdutl' '.')
% 'rm' '-f' 'bin/etcdctl'
% (cd etcdctl && 'env' 'GO_BUILD_FLAGS=' 'CGO_ENABLED=0' 'GO_BUILD_FLAGS=' 'GOOS=linux' 'GOARCH=amd64' 'go' 'build' '-trimpath' '-installsuffix=cgo' '-ldflags=-X=go.etcd.io/etcd/api/v3/version.GitSHA=9a55333-FAILPOINTS' '-o=../bin/etcdctl' '.')
SUCCESS: etcd_build (GOARCH=amd64)
etcd $ cd  /tmp/etcd-v3.5.15-failpoints/tools/mod/
mod $ go list -m -f '{{.Version}}' go.etcd.io/gofail
v0.1.0

ahrtr · 2024-08-02T10:03:57Z

/tmp/etcd-v3.5.15-failpoints/bin

It makes sense now. Your previous example was based on /tmp/etcd-release-3.5-failpoints/bin.

ahrtr

LGTM

Thanks

fuweid

LGTM

k8s-ci-robot · 2024-08-02T12:22:36Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: ahrtr, fuweid, henrybear327, serathius

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

~~OWNERS~~ [ahrtr,serathius]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

Ensure proper gofail package version in robustness tests

44b6c03

Signed-off-by: Marek Siarkowicz <[email protected]>

serathius force-pushed the robustness-gofail-branch branch from 6b48ee2 to 44b6c03 Compare August 2, 2024 07:25

k8s-ci-robot added area/robustness-testing area/testing approved labels Aug 2, 2024

k8s-ci-robot added size/XS labels Aug 2, 2024

henrybear327 reviewed Aug 2, 2024

View reviewed changes

ahrtr approved these changes Aug 2, 2024

View reviewed changes

fuweid approved these changes Aug 2, 2024

View reviewed changes

henrybear327 approved these changes Aug 2, 2024

View reviewed changes

serathius merged commit e758ffc into etcd-io:main Aug 2, 2024
77 checks passed

henrybear327 mentioned this pull request Aug 28, 2024

Failpoint raftAfterSave=sleep(1s) is flaking in robustness test #18240

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Ensure proper gofail package version in robustness tests #18397

Ensure proper gofail package version in robustness tests #18397

serathius commented Aug 2, 2024 •

edited

Loading

serathius commented Aug 2, 2024

henrybear327 Aug 2, 2024

serathius Aug 2, 2024

henrybear327 Aug 2, 2024

ahrtr Aug 2, 2024

ahrtr commented Aug 2, 2024 •

edited

Loading

serathius commented Aug 2, 2024

ahrtr commented Aug 2, 2024

ahrtr commented Aug 2, 2024

serathius commented Aug 2, 2024 •

edited

Loading

ahrtr commented Aug 2, 2024

ahrtr left a comment

fuweid left a comment

k8s-ci-robot commented Aug 2, 2024

	cd ./server && go get go.etcd.io/gofail@"${GOFAIL_VERSION}"
	cd ../etcdutl && go get go.etcd.io/gofail@"${GOFAIL_VERSION}"
	cd ../etcdctl && go get go.etcd.io/gofail@"${GOFAIL_VERSION}"
	cd ../tests && go get go.etcd.io/gofail@"${GOFAIL_VERSION}"

Ensure proper gofail package version in robustness tests #18397

Ensure proper gofail package version in robustness tests #18397

Conversation

serathius commented Aug 2, 2024 • edited Loading

serathius commented Aug 2, 2024

henrybear327 Aug 2, 2024

Choose a reason for hiding this comment

serathius Aug 2, 2024

Choose a reason for hiding this comment

henrybear327 Aug 2, 2024

Choose a reason for hiding this comment

ahrtr Aug 2, 2024

Choose a reason for hiding this comment

ahrtr commented Aug 2, 2024 • edited Loading

serathius commented Aug 2, 2024

ahrtr commented Aug 2, 2024

ahrtr commented Aug 2, 2024

serathius commented Aug 2, 2024 • edited Loading

ahrtr commented Aug 2, 2024

ahrtr left a comment

Choose a reason for hiding this comment

fuweid left a comment

Choose a reason for hiding this comment

k8s-ci-robot commented Aug 2, 2024

serathius commented Aug 2, 2024 •

edited

Loading

ahrtr commented Aug 2, 2024 •

edited

Loading

serathius commented Aug 2, 2024 •

edited

Loading