We want the KV to support recovering from backend to avoid
additional pointer swap. Or we have to do coordination between
etcdserver and API layer, since API layer might have access to
kv pointer and use a closed kv.
Provides two implementations of Recorder-- one that is non-blocking
like the original version and one that provides a blocking channel
to avoid busy waiting or racing in tests when no other synchronization
is available.
In this case, we know we are waiting for an action happened on
storage. We can do a busy wait instead of calling waitSchedule.
The test previously failed on CI with no observed actions.
When a slow follower receives the snapshot sent from the leader, it
should rename the snapshot file to the default KV file path, and
restore KV snapshot.
Have tested it manually and it works pretty well.
When snapshot store requests raft snapshot from etcdserver apply loop,
it may block on the channel for some time, or wait some time for KV to
snapshot. This is unexpected because raft state machine should be unblocked.
Even worse, this block may lead to deadlock:
1. raft state machine waits on getting snapshot from raft memory storage
2. raft memory storage waits snapshot store to get snapshot
3. snapshot store requests raft snapshot from apply loop
4. apply loop is applying entries, and waits raftNode loop to finish
messages sending
5. raftNode loop waits peer loop in Transport to send out messages
6. peer loop in Transport waits for raft state machine to process message
Fix it by changing the logic of getSnap to be asynchronously creation.
Use snapshotSender to send v3 snapshot message. It puts raft snapshot
message and v3 snapshot into request body, then sends it to the target peer.
When it receives http.StatusNoContent, it knows the message has been
received and processed successfully.
As receiver, snapHandler saves v3 snapshot and then processes the raft snapshot
message, then respond with http.StatusNoContent.
rafthttp has different requirements for connections created by the
transport for different usage, and this is hard to achieve when giving
one http.RoundTripper. Pass into pkg the data needed to build transport
now, and let rafthttp build its own transports.
time.AfterFunc() creates its own goroutine and calls the callback
function in the goroutine. It can cause datarace like the problem
fixed in the commit de1a16e0f1 . This
commit also fixes the potential dataraces of tests in
etcdserver/server_test.go .
snapshotStore is the store of snapshot, and it supports to get latest snapshot
and save incoming snapshot.
raftStorage supports to get latest snapshot when v3demo is open.
Like the commit 6974fc63ed, this commit lets etcdserver forbid
removing started member if quorum cannot be preserved after
reconfiguration if the option -strict-reconfig-check is passed to
etcd. The removal can cause deadlock if unstarted members have wrong
peer URLs.
Current membership changing functionality of etcd seems to have a
problem which can cause deadlock.
How to produce:
1. construct N node cluster
2. add N new nodes with etcdctl member add, without starting the new members
What happens:
After finishing add N nodes, a total number of the cluster becomes 2 *
N and a quorum number of the cluster becomes N + 1. It means
membership change requires at least N + 1 nodes because Raft treats
membership information in its log like other ordinal log append
requests.
Assume the peer URLs of the added nodes are wrong because of miss
operation or bugs in wrapping program which launch etcd. In such a
case, both of adding and removing members are impossible because the
quorum isn't preserved. Of course ordinal requests cannot be
served. The cluster would seem to be deadlock.
Of course, the best practice of adding new nodes is adding one node
and let the node start one by one. However, the effect of this problem
is so serious. I think preventing the problem forcibly would be
valuable.
Solution:
This patch lets etcd forbid adding a new node if the operation changes
quorum and the number of changed quorum is larger than a number of
running nodes. If etcd is launched with a newly added option
-strict-reconfig-check, the checking logic is activated. If the option
isn't passed, default behavior of reconfig is kept.
Fixes https://github.com/coreos/etcd/issues/3477
It specifies request timeout error possibly caused by connection lost,
and print out better log for user to understand.
It handles two cases:
1. the leader cannot connect to majority of cluster.
2. the connection between follower and leader is down for a while,
and it losts proposals.
log format:
```
20:04:19 etcd3 | 2015-08-25 20:04:19.368126 E | etcdhttp: etcdserver:
request timed out, possibly due to connection lost
20:04:19 etcd3 | 2015-08-25 20:04:19.368227 E | etcdhttp: etcdserver:
request timed out, possibly due to connection lost
```
It uses heartbeat interval and election timeout to estimate the
commit timeout for internal requests.
This PR helps etcd survive under high roundtrip-time environment,
e.g., globally-deployed cluster.
Before this PR, it may error like this:
```
--- FAIL: TestUpdateMember-2 (0.00s)
server_test.go:950: action =
[{ApplyConfChange:ConfChangeUpdateNode []}
{ProposeConfChange:ConfChangeUpdateNode []}], want
[{ProposeConfChange:ConfChangeUpdateNode []}
{ApplyConfChange:ConfChangeUpdateNode []}]
```
This fixes the test by recording the proposal event in time.
ForceGosched() performs bad when GOMAXPROCS>1. When GOMAXPROCS=1, it
could promise that other goroutines run long enough
because it always yield the processor to other goroutines. But it cannot
yield processor to goroutine running on other processors. So when
GOMAXPROCS>1, the yield may finish when goroutine on the other
processor just runs for little time.
Here is a test to confirm the case:
```
package main
import (
"fmt"
"runtime"
"testing"
)
func ForceGosched() {
// possibility enough to sched up to 10 go routines.
for i := 0; i < 10000; i++ {
runtime.Gosched()
}
}
var d int
func loop(c chan struct{}) {
for {
select {
case <-c:
for i := 0; i < 1000; i++ {
fmt.Sprintf("come to time %d", i)
}
d++
}
}
}
func TestLoop(t *testing.T) {
c := make(chan struct{}, 1)
go loop(c)
c <- struct{}{}
ForceGosched()
if d != 1 {
t.Fatal("d is not incremented")
}
}
```
`go test -v -race` runs well, but `GOMAXPROCS=2 go test -v -race` fails.
Change the functionality to waiting for schedule to happen.
After this PR, only cluster's interface Cluster is exposed, which makes
code much cleaner. And it avoids external packages to rely on cluster
struct in the future.
1. Persist the cluster version change through raft. When the member is restarted, it can recover
the previous known decided cluster version.
2. When there is a new leader, it is forced to do a version checking immediately. This helps to
update the first cluster version fast.
Add remotes to rafthttp, who help newly joined members catch up the
progress of the cluster. It supports basic message sending to remote, and
has no stream connection for simplicity. remotes will not be used
after the latest peers have been added into rafthttp.
Encoding store into json snapshot has quite high CPU cost. And it
will block for a while. This commit makes the encoding process non-
blocking by running it in another go-routine.
raft relies on the link layer to report the status of the sent snapshot.
If the snapshot is still sending, the replication to that remote peer will
be paused. If the snapshot finish sending, the replication will begin
optimistically after electionTimeout. If the snapshot fails, raft will
try to resend it.
WAL should control the cut logic itself. We want to do falloc to
per allocate the space for a segmented wal file at the beginning
and cut it when it size reaches the limit.
Anthony Romano