(Node) Enhance raft test codes #2289
Conversation
📝 WalkthroughWalkthroughThe changes involve significant modifications to the Raft consensus algorithm implementation in the Changes
Sequence Diagram(s)sequenceDiagram
participant Client
participant RaftNode
participant Blacklist
Client->>RaftNode: Send Message
RaftNode->>Blacklist: Check if sender is blacklisted
alt Sender is blacklisted
Blacklist-->>RaftNode: Return true
RaftNode-->>Client: Ignore message
else Sender is not blacklisted
Blacklist-->>RaftNode: Return false
RaftNode-->>Client: Process message
end
Thank you for using CodeRabbit. We offer it for free to the OSS community and would appreciate your support in helping us grow. If you find it useful, would you consider giving us a shout-out on your favorite social media? 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments. CodeRabbit Commands (Invoked using PR comments)
Other keywords and placeholders
CodeRabbit Configuration File (
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 8
🧹 Outside diff range and nitpick comments (6)
node/pkg/raft/raft_test.go (6)
72-79: Handle potential errors inisolateNodefunctionIn the
isolateNodefunction, you are callingentry.addBlacklist(node.GetHostId())without checking for errors. Consider handling any potential errors returned byaddBlacklistto ensure robustness and facilitate debugging.Apply this diff to handle errors:
for _, entry := range nodes { if entry.GetHostId() == node.GetHostId() { continue } - entry.addBlacklist(node.GetHostId()) + if err := entry.addBlacklist(node.GetHostId()); err != nil { + log.Error().Err(err).Msgf("Failed to isolate node %s", node.GetHostId()) + } }
81-88: Handle potential errors inrecoverNodefunctionSimilarly, in the
recoverNodefunction, you're callingentry.removeBlacklist(node.GetHostId())without error handling. It's good practice to handle or log any errors to assist with troubleshooting.Apply this diff to handle errors:
for _, entry := range nodes { if entry.GetHostId() == node.GetHostId() { continue } - entry.removeBlacklist(node.GetHostId()) + if err := entry.removeBlacklist(node.GetHostId()); err != nil { + log.Error().Err(err).Msgf("Failed to recover node %s", node.GetHostId()) + } }
90-90: Update function comments to reflect new parameterThe
setupfunction signature now includesnumOfNodesas a parameter. Updating the function's comment to include this parameter will improve code clarity and maintainability.
161-162: Update documentation forsetupRaftClusterWith the addition of
numOfNodesto thesetupRaftClusterfunction, ensure that any associated comments or documentation are updated to reflect this change for consistency and clarity.
310-313: Add cleanup fornewNode.Topicin testsIn your test case, after creating
newNode, you defernewNode.Host.Close()but don't closenewNode.Topic. To ensure all resources are released, consider closing the topic as well.Apply this diff:
defer newNode.Host.Close() +defer newNode.Topic.Close()
865-866: Ensure consistency in log messages across testsIn
TestRaft_MessageDelay_2_nodes, you correctly log "Ready to test, setup 2 nodes". Review other test functions to ensure all log messages accurately reflect the number of nodes involved, enhancing clarity during test executions.
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
📒 Files selected for processing (3)
- node/pkg/raft/raft.go (6 hunks)
- node/pkg/raft/raft_test.go (10 hunks)
- node/pkg/raft/types.go (1 hunks)
🔇 Additional comments (4)
node/pkg/raft/types.go (1)
77-79: Consider separating test-specific fields and clarify changes
The addition of test-specific fields (
blacklistandisDelayedNode) to the mainRaftstruct might not be ideal:
- It could lead to confusion in production code.
- It might unnecessarily increase the memory footprint in non-test scenarios.
Consider moving these fields to a separate test-specific struct or using dependency injection for testing.
Please clarify the purpose and usage of the new fields:
- What is the
blacklistused for in the context of Raft testing?- How is the
isDelayedNodeflag utilized in test scenarios?The AI summary mentions the removal of the
Resign chan interface{}field. Could you please explain the implications of this removal? How will the node step down from leadership without this channel?To verify the usage and impact of these changes, please run the following script:
This script will help us understand how these new fields are used and if there are any remaining references to the removed
Resignchannel.✅ Verification successful
Verification Complete: Changes Confirmed
Test-Specific Fields:
- The
blacklistandisDelayedNodefields are utilized withinraft.goandtypes.gofor testing purposes, aligning with their intended use.Resign Channel Removal:
- The
Resign chan interface{}has been successfully removed, with no remaining references found in the codebase.🏁 Scripts executed
The following scripts were executed for the analysis:
Script:
#!/bin/bash # Description: Check for usage of new fields and removed Resign channel # Check for blacklist usage echo "Checking blacklist usage:" rg --type go 'blacklist' ./node/pkg/raft # Check for isDelayedNode usage echo "Checking isDelayedNode usage:" rg --type go 'isDelayedNode' ./node/pkg/raft # Check for Resign channel usage (to verify removal impact) echo "Checking Resign channel usage:" rg --type go 'Resign\s+chan\s+interface{}' ./node/pkg/raftLength of output: 1113
Script:
#!/bin/bash # Description: Check for usage of new fields and removed Resign channel # Check for blacklist usage echo "Checking blacklist usage:" rg --type go 'blacklist' ./node/pkg/raft # Check for isDelayedNode usage echo "Checking isDelayedNode usage:" rg --type go 'isDelayedNode' ./node/pkg/raft # Check for Resign channel usage (corrected regex) echo "Checking Resign channel usage:" rg --type go 'Resign\s+chan\s+interface\{\}' ./node/pkg/raftLength of output: 979
node/pkg/raft/raft_test.go (3)
59-61: Verify term comparison logic when only one node is presentIn the
allNodesHaveSameTermMinfunction, when there's only one node, you're checking if its current term is greater thanmin. Ensure that this condition correctly represents the intended behavior in scenarios with a single node, and that it doesn't cause any unintended side effects in your tests.
799-801: Verify network partition behavior in a two-node clusterWith only two nodes in
TestRaft_NetworkPartitionAndRecovery_2_nodes, a network partition can lead to a split-brain scenario where neither node can form a majority. Verify that your Raft implementation correctly handles this case, and the test adequately assesses the cluster's behavior under partition.
743-743: Re-evaluate the necessity of the assertion onnewListlengthIn
TestRaft_LeaderDisconnect_2_nodes, you assert thatlen(newList)equals 1 after removing the old leader. Confirm that this assertion is required for the test and accurately reflects the expected number of nodes remaining in the cluster.
| blacklist: make(map[string]struct{}), | ||
| isDelayedNode: false, |
There was a problem hiding this comment.
Potential Data Races on 'blacklist' and 'isDelayedNode'
The fields blacklist and isDelayedNode are accessed by multiple goroutines without adequate synchronization, which could lead to data races and panics.
To fix this issue, ensure that all accesses to blacklist and isDelayedNode are protected by the mutex r.Mutex. For example, modify the handleMessage and PublishMessage methods to acquire the lock when accessing these fields.
| func (r *Raft) addBlacklist(id string) { | ||
| // used for testing purposes | ||
| r.Mutex.Lock() | ||
| defer r.Mutex.Unlock() | ||
| r.blacklist[id] = struct{}{} | ||
| } | ||
|
|
||
| func (r *Raft) removeBlacklist(id string) { | ||
| // used for testing purposes | ||
| r.Mutex.Lock() | ||
| defer r.Mutex.Unlock() | ||
| delete(r.blacklist, id) | ||
| } | ||
|
|
||
| func (r *Raft) setDelayedNode(isDelayed bool) { | ||
| // used for testing purposes | ||
| r.Mutex.Lock() | ||
| defer r.Mutex.Unlock() | ||
| r.isDelayedNode = isDelayed | ||
| } |
There was a problem hiding this comment.
🛠️ Refactor suggestion
Embedding Testing Code into Production Code
The methods addBlacklist, removeBlacklist, and setDelayedNode are added for testing purposes but are part of the production codebase. Including testing utilities in production code can lead to maintenance challenges and potential misuse.
Consider refactoring to separate testing utilities from production code. Options include:
- Moving these methods to a build tag dedicated to testing.
- Using interfaces or dependency injection to mock behaviors in tests without altering the production codebase.
- Placing testing code within a
_test.gofile to ensure it's only included during testing.
This separation enhances code maintainability and prevents testing code from impacting production.
| if r.isDelayedNode { | ||
| time.Sleep(500 * time.Millisecond) | ||
| } |
There was a problem hiding this comment.
Unsafe Access to 'isDelayedNode' in 'PublishMessage'
The isDelayedNode flag is accessed without synchronization in PublishMessage, which can result in data races when setDelayedNode is called concurrently.
Apply this diff to synchronize access to isDelayedNode:
func (r *Raft) PublishMessage(ctx context.Context, msg Message) error {
data, err := json.Marshal(msg)
if err != nil {
return err
}
+ r.Mutex.Lock()
if r.isDelayedNode {
time.Sleep(500 * time.Millisecond)
}
+ r.Mutex.Unlock()
return r.Topic.Publish(ctx, data)
}Ensure all accesses to isDelayedNode are properly synchronized.
📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| if r.isDelayedNode { | |
| time.Sleep(500 * time.Millisecond) | |
| } | |
| r.Mutex.Lock() | |
| if r.isDelayedNode { | |
| time.Sleep(500 * time.Millisecond) | |
| } | |
| r.Mutex.Unlock() |
| if r.Role != Leader { | ||
| r.startElectionTimer() | ||
| } |
There was a problem hiding this comment.
Unnecessary Election Timer Reset in Follower State
In handleHeartbeat, the election timer is restarted even if the node is not the leader. Restarting the election timer upon receiving a heartbeat as a follower may not be necessary and could interfere with the expected timing of elections.
Review the logic to ensure the election timer is only restarted when appropriate. If the node is a follower and receives a heartbeat, the election timer should already be managed correctly. Consider modifying the code as follows:
func (r *Raft) handleHeartbeat(msg Message) error {
// existing code...
- if r.Role != Leader {
+ if r.Role == Candidate {
r.startElectionTimer()
}
// existing code...
}This change ensures that only candidates reset their election timers upon receiving a heartbeat, aligning with typical Raft protocol behavior.
📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| if r.Role != Leader { | |
| r.startElectionTimer() | |
| } | |
| if r.Role == Candidate { | |
| r.startElectionTimer() | |
| } |
| r.Role = Follower | ||
| r.LeaderID = "" | ||
|
|
||
| r.HeartbeatTicker.Stop() | ||
| r.LeaderJobTicker.Stop() | ||
|
|
||
| r.startElectionTimer() |
There was a problem hiding this comment.
Nil Pointer Dereference Risk when Stopping Tickers
In ResignLeader, calling Stop() on r.HeartbeatTicker and r.LeaderJobTicker without checking if they are initialized may lead to nil pointer dereference if these tickers are nil.
Apply this diff to add nil checks before stopping the tickers:
func (r *Raft) ResignLeader() {
r.Role = Follower
r.LeaderID = ""
- r.HeartbeatTicker.Stop()
- r.LeaderJobTicker.Stop()
+ if r.HeartbeatTicker != nil {
+ r.HeartbeatTicker.Stop()
+ }
+ if r.LeaderJobTicker != nil {
+ r.LeaderJobTicker.Stop()
+ }
r.startElectionTimer()
}This ensures that Stop() is only called on initialized tickers.
📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| r.Role = Follower | |
| r.LeaderID = "" | |
| r.HeartbeatTicker.Stop() | |
| r.LeaderJobTicker.Stop() | |
| r.startElectionTimer() | |
| func (r *Raft) ResignLeader() { | |
| r.Role = Follower | |
| r.LeaderID = "" | |
| if r.HeartbeatTicker != nil { | |
| r.HeartbeatTicker.Stop() | |
| } | |
| if r.LeaderJobTicker != nil { | |
| r.LeaderJobTicker.Stop() | |
| } | |
| r.startElectionTimer() | |
| } |
| if _, ok := r.blacklist[msg.SentFrom]; ok { | ||
| // blacklist for testing to simulate network partition | ||
| return nil | ||
| } | ||
|
|
There was a problem hiding this comment.
Unsafe Access to 'blacklist' in 'handleMessage'
In the handleMessage method, the access to r.blacklist is not protected by a mutex, leading to potential concurrent map read and write, which can cause runtime panics.
Apply this diff to safely access blacklist:
func (r *Raft) handleMessage(ctx context.Context, msg Message) error {
+ r.Mutex.Lock()
if _, ok := r.blacklist[msg.SentFrom]; ok {
// blacklist for testing to simulate network partition
+ r.Mutex.Unlock()
return nil
}
+ r.Mutex.Unlock()Alternatively, create a helper function that safely checks the blacklist with proper synchronization.
Committable suggestion was skipped due to low confidence.
| log.Info().Msg("Ready to test, setup 3 nodes") | ||
|
|
There was a problem hiding this comment.
Correct log message to match the test setup
In TestRaft_NetworkPartitionAndRecovery_2_nodes, the log message states "Ready to test, setup 3 nodes", but the test is set up with 2 nodes. Updating the log message will prevent confusion.
Apply this diff to fix the log message:
- log.Info().Msg("Ready to test, setup 3 nodes")
+ log.Info().Msg("Ready to test, setup 2 nodes")📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| log.Info().Msg("Ready to test, setup 3 nodes") | |
| log.Info().Msg("Ready to test, setup 2 nodes") |
| func joinNewNode(ctx context.Context, testItems *TestItems) (*Raft, error) { | ||
| newHost, err := libp2psetup.NewHost(ctx) | ||
| if err != nil { | ||
| return nil, err | ||
| } | ||
|
|
||
| newPs, err := libp2psetup.MakePubsub(ctx, newHost) | ||
| if err != nil { | ||
| return nil, err | ||
| } | ||
|
|
||
| for _, host := range testItems.Hosts { | ||
| newHost.Connect(ctx, peer.AddrInfo{ID: host.ID(), Addrs: host.Addrs()}) | ||
| } | ||
|
|
||
| newTopic, err := newPs.Join(topicString) | ||
| if err != nil { | ||
| return nil, err | ||
| } | ||
|
|
||
| newNode := NewRaftNode(newHost, newPs, newTopic, 100, time.Second) | ||
| go newNode.Run(ctx) | ||
| return newNode, nil | ||
| } |
There was a problem hiding this comment.
Ensure proper cleanup of resources in joinNewNode
The joinNewNode function creates a new node and starts it but doesn't include cleanup logic for the node's resources. To prevent resource leaks, make sure to close the host and topic when they are no longer needed.
Apply this diff to add cleanup:
func joinNewNode(ctx context.Context, testItems *TestItems) (*Raft, error) {
// ... existing code ...
newNode := NewRaftNode(newHost, newPs, newTopic, 100, time.Second)
go newNode.Run(ctx)
- return newNode, nil
+ return newNode, nil
+}
+
+// In the calling test function, add defer statements after successful node creation:
+defer func() {
+ newNode.Topic.Close()
+ newNode.Host.Close()
+}()Committable suggestion was skipped due to low confidence.
Description
Minor updates
Type of change
Please delete options that are not relevant.
Checklist before requesting a review
Deployment