-
Notifications
You must be signed in to change notification settings - Fork 169
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
2. Add random fault tests 3. Update readme to include instructions on validation.sh Signed-off-by: Joshua Zhang <[email protected]>
- Loading branch information
1 parent
7f35117
commit a66c3e6
Showing
19 changed files
with
5,223 additions
and
2,923 deletions.
There are no files selected for viewing
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
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
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
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
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,260 @@ | ||
| package rafttest | ||
|
|
||
| import ( | ||
| "context" | ||
| "fmt" | ||
| "time" | ||
|
|
||
| "go.etcd.io/raft/v3" | ||
| "go.etcd.io/raft/v3/raftpb" | ||
| ) | ||
|
|
||
| type clusterConfig struct { | ||
| size int | ||
| traceLogger raft.TraceLogger | ||
| tickInterval time.Duration | ||
| } | ||
|
|
||
| type endpoint struct { | ||
| index int | ||
| node *node | ||
| } | ||
|
|
||
| type getEndpoint struct { | ||
| i int | ||
| c chan endpoint | ||
| } | ||
|
|
||
| type cluster struct { | ||
| nodes map[uint64]*node | ||
| network *raftNetwork | ||
|
|
||
| stopc chan struct{} | ||
| removec chan uint64 | ||
| addc chan uint64 | ||
| getc chan getEndpoint | ||
| faultc chan func(*cluster) | ||
|
|
||
| traceLogger raft.TraceLogger | ||
| tickInterval time.Duration | ||
| } | ||
|
|
||
| func newCluster(c clusterConfig) *cluster { | ||
| tickInterval := c.tickInterval | ||
| if tickInterval == 0 { | ||
| tickInterval = 100 * time.Millisecond | ||
| } | ||
| ids := make([]uint64, c.size) | ||
| peers := make([]raft.Peer, c.size) | ||
| for i := 0; i < c.size; i++ { | ||
| peers[i].ID = uint64(i + 1) | ||
| ids[i] = peers[i].ID | ||
| } | ||
| network := newRaftNetwork(ids...) | ||
| nodes := make(map[uint64]*node, c.size) | ||
| for i := 0; i < c.size; i++ { | ||
| nodes[uint64(i+1)] = startNodeWithConfig(nodeConfig{ | ||
| id: uint64(i + 1), | ||
| peers: peers, | ||
| iface: network.nodeNetwork(uint64(i + 1)), | ||
| traceLogger: c.traceLogger, | ||
| tickInterval: tickInterval, | ||
| }) | ||
| } | ||
|
|
||
| cl := &cluster{ | ||
| nodes: nodes, | ||
| network: network, | ||
| stopc: make(chan struct{}), | ||
| removec: make(chan uint64), | ||
| addc: make(chan uint64), | ||
| getc: make(chan getEndpoint), | ||
| faultc: make(chan func(*cluster)), | ||
| traceLogger: c.traceLogger, | ||
| tickInterval: tickInterval, | ||
| } | ||
|
|
||
| cl.waitLeader() | ||
|
|
||
| go cl.mgmtLoop() | ||
|
|
||
| return cl | ||
| } | ||
|
|
||
| func (cl *cluster) mgmtLoop() { | ||
| peers := []raft.Peer{} | ||
| for _, n := range cl.nodes { | ||
| peers = append(peers, raft.Peer{ID: n.id, Context: []byte{}}) | ||
| } | ||
| for { | ||
| select { | ||
| case id := <-cl.addc: | ||
| cl.network.changeFace(id, true) | ||
| peers = append(peers, raft.Peer{ID: id, Context: []byte{}}) | ||
| node := startNodeWithConfig(nodeConfig{ | ||
| id: id, | ||
| peers: nil, | ||
| iface: cl.network.nodeNetwork(id), | ||
| traceLogger: cl.traceLogger, | ||
| tickInterval: cl.tickInterval, | ||
| }) | ||
| cl.nodes[id] = node | ||
| case id := <-cl.removec: | ||
| cl.network.changeFace(id, false) | ||
| cl.nodes[id].stop() | ||
| delete(cl.nodes, id) | ||
| for i, p := range peers { | ||
| if p.ID == id { | ||
| peers = append(peers[:i], peers[i+1:]...) | ||
| break | ||
| } | ||
| } | ||
| case gn := <-cl.getc: | ||
| i := gn.i % len(peers) | ||
| nid := peers[i].ID | ||
| gn.c <- endpoint{index: i, node: cl.nodes[nid]} | ||
| case <-cl.stopc: | ||
| for _, n := range cl.nodes { | ||
| n.stop() | ||
| } | ||
| close(cl.stopc) | ||
| return | ||
| case f := <-cl.faultc: | ||
| cl.network.clearFault() | ||
| for _, n := range cl.nodes { | ||
| if n.stopped { | ||
| n.restart() | ||
| } | ||
| } | ||
| f(cl) | ||
| } | ||
| } | ||
| } | ||
|
|
||
| func (cl *cluster) stop() { | ||
| cl.stopc <- struct{}{} | ||
| <-cl.stopc | ||
| } | ||
|
|
||
| func (cl *cluster) removeNode(id uint64) { | ||
| cl.removec <- id | ||
| } | ||
|
|
||
| func (cl *cluster) addNode(id uint64) { | ||
| cl.addc <- id | ||
| } | ||
|
|
||
| func (cl *cluster) newClient() *client { | ||
| return &client{cluster: cl, epc: make(chan endpoint)} | ||
| } | ||
|
|
||
| func (cl *cluster) waitLeader() uint64 { | ||
| var l map[uint64]struct{} | ||
| var lindex uint64 | ||
|
|
||
| for { | ||
| l = make(map[uint64]struct{}) | ||
|
|
||
| for i, n := range cl.nodes { | ||
| lead := n.Status().SoftState.Lead | ||
| if lead != 0 { | ||
| l[lead] = struct{}{} | ||
| if n.id == lead { | ||
| lindex = i | ||
| } | ||
| } | ||
| } | ||
|
|
||
| if len(l) == 1 { | ||
| return lindex | ||
| } | ||
| } | ||
| } | ||
|
|
||
| type client struct { | ||
| cluster *cluster | ||
| epi int | ||
| epc chan endpoint | ||
| } | ||
|
|
||
| func (cl *client) propose(ctx context.Context, data []byte) error { | ||
| ep := cl.getEndpoint() | ||
| return ep.Propose(ctx, data) | ||
| } | ||
|
|
||
| func (cl *client) addNode(ctx context.Context, n uint64) error { | ||
| change := raftpb.ConfChangeSingle{ | ||
| Type: raftpb.ConfChangeAddNode, | ||
| NodeID: n, | ||
| } | ||
| cc := raftpb.ConfChangeV2{ | ||
| Transition: 0, | ||
| Changes: []raftpb.ConfChangeSingle{change}, | ||
| Context: []byte{}, | ||
| } | ||
|
|
||
| ep := cl.getEndpoint() | ||
| if err := ep.ProposeConfChange(ctx, cc); err != nil { | ||
| return err | ||
| } | ||
|
|
||
| toc := time.After(cl.cluster.tickInterval * 50) | ||
| for { | ||
| select { | ||
| case <-toc: | ||
| return fmt.Errorf("addNode timeout") | ||
| default: | ||
| } | ||
| st := ep.Status() | ||
| if _, exist := st.Config.Voters[0][n]; exist { | ||
| break | ||
| } | ||
| } | ||
| cl.cluster.addNode(n) | ||
| return nil | ||
| } | ||
|
|
||
| func (cl *client) removeNode(ctx context.Context, n uint64) error { | ||
| change := raftpb.ConfChangeSingle{ | ||
| Type: raftpb.ConfChangeRemoveNode, | ||
| NodeID: n, | ||
| } | ||
| cc := raftpb.ConfChangeV2{ | ||
| Transition: 0, | ||
| Changes: []raftpb.ConfChangeSingle{change}, | ||
| Context: []byte{}, | ||
| } | ||
|
|
||
| ep := cl.getEndpoint() | ||
| if err := ep.ProposeConfChange(ctx, cc); err != nil { | ||
| return err | ||
| } | ||
|
|
||
| toc := time.After(cl.cluster.tickInterval * 50) | ||
| for { | ||
| select { | ||
| case <-toc: | ||
| return fmt.Errorf("removeNode timeout") | ||
| default: | ||
| } | ||
| st := ep.Status() | ||
| if _, exist := st.Config.Voters[0][n]; !exist { | ||
| break | ||
| } | ||
| } | ||
|
|
||
| cl.cluster.removeNode(n) | ||
| return nil | ||
| } | ||
|
|
||
| func (cl *client) getEndpoint() *node { | ||
| // round robin | ||
| ge := getEndpoint{ | ||
| i: cl.epi + 1, | ||
| c: cl.epc, | ||
| } | ||
| cl.cluster.getc <- ge | ||
| ep := <-ge.c | ||
| cl.epi = ep.index | ||
| return ep.node | ||
| } |
Oops, something went wrong.