Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

fix(namespace): didn't reload namespaces from DB after the full sync #2527

Merged
merged 2 commits into from
Sep 8, 2024
Merged

fix(namespace): didn't reload namespaces from DB after the full sync #2527

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
1dd0df0
Fix didn't reload namespaces from DB after the full sync
git-hulk Sep 8, 2024
003f424
Merge branch 'unstable' into fix/ns-replication-not-work
git-hulk Sep 8, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions src/cluster/replication.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -732,6 +732,13 @@ ReplicationThread::CBState ReplicationThread::fullSyncReadCB(bufferevent *bev) {
LOG(INFO) << "[replication] Succeeded restoring the backup, fullsync was finish";
post_fullsync_cb_();

// It needs to reload namespaces from DB after the full sync is done,
// or namespaces are not visible in the replica.
s = srv_->GetNamespace()->LoadAndRewrite();
if (!s.IsOK()) {
LOG(ERROR) << "[replication] Failed to load and rewrite namespace: " << s.Msg();
}

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Extract a PostFullSync function, move it to this function?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We have post_fullsync_cb now, but it's used for cleaning resources if failed to do the full sync. So it should be not good to put this inside that.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Isn't it a callback function after successfully loading the data?

kvrocks/src/server/server.cc

Lines 284 to 290 in 2836bf0

auto s = replication_thread_->Start([this]() { PrepareRestoreDB(); },
[this]() {
this->is_loading_ = false;
if (auto s = task_runner_.Start(); !s) {
LOG(WARNING) << "Failed to start task runner: " << s.Msg();
}
});

Copy link
Member Author

@git-hulk git-hulk Sep 8, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No indeed, it should be used for resetting resources corresponding to pre_fullsync_cb. That said, it must be called once if the pre_fullsync_cb was invoked. I will submit another PR to fix #2511 which is caused by that didn't call post_fullsync_cb while failing to restore DB.

// Switch to psync state machine again
psync_steps_.Start();
return CBState::QUIT;
Expand Down
3 changes: 2 additions & 1 deletion src/server/namespace.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,7 +54,8 @@ bool Namespace::IsAllowModify() const {
Status Namespace::loadFromDB(std::map<std::string, std::string>* db_tokens) const {
std::string value;
engine::Context ctx(storage_);
auto s = storage_->Get(ctx, ctx.GetReadOptions(), cf_, kNamespaceDBKey, &value);
auto cf = storage_->GetCFHandle(ColumnFamilyID::Propagate);
auto s = storage_->Get(ctx, ctx.GetReadOptions(), cf, kNamespaceDBKey, &value);
if (!s.ok()) {
if (s.IsNotFound()) return Status::OK();
return {Status::NotOK, s.ToString()};
Expand Down
4 changes: 1 addition & 3 deletions src/server/namespace.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,8 +26,7 @@ constexpr const char *kNamespaceDBKey = "__namespace_keys__";

class Namespace {
public:
explicit Namespace(engine::Storage *storage)
: storage_(storage), cf_(storage_->GetCFHandle(ColumnFamilyID::Propagate)) {}
explicit Namespace(engine::Storage *storage) : storage_(storage) {}

~Namespace() = default;
Namespace(const Namespace &) = delete;
Expand All @@ -45,7 +44,6 @@ class Namespace {

private:
engine::Storage *storage_;
rocksdb::ColumnFamilyHandle *cf_ = nullptr;

std::shared_mutex tokens_mu_;
// mapping from token to namespace name
Expand Down
39 changes: 39 additions & 0 deletions tests/gocase/unit/namespace/namespace_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,8 @@ package namespace

import (
"context"
"fmt"
"strings"
"sync"
"testing"
"time"
Expand Down Expand Up @@ -252,6 +254,43 @@ func TestNamespaceReplicate(t *testing.T) {
})
}

func TestNamespaceReplicateWithFullSync(t *testing.T) {
config := map[string]string{
"rocksdb.write_buffer_size": "4",
"rocksdb.target_file_size_base": "16",
"rocksdb.max_write_buffer_number": "1",
"rocksdb.wal_ttl_seconds": "0",
"rocksdb.wal_size_limit_mb": "0",
"repl-namespace-enabled": "yes",
"requirepass": "123",
"masterauth": "123",
}
master := util.StartServer(t, config)
defer master.Close()
masterClient := master.NewClientWithOption(&redis.Options{Password: "123"})
defer func() { require.NoError(t, masterClient.Close()) }()

slave := util.StartServer(t, config)
defer slave.Close()
slaveClient := slave.NewClientWithOption(&redis.Options{Password: "123"})
defer func() { require.NoError(t, slaveClient.Close()) }()

ctx := context.Background()
value := strings.Repeat("a", 128*1024)
for i := 0; i < 1024; i++ {
require.NoError(t, masterClient.Set(ctx, fmt.Sprintf("key%d", i), value, 0).Err())
}
require.NoError(t, masterClient.Do(ctx, "NAMESPACE", "ADD", "foo", "bar").Err())

util.SlaveOf(t, slaveClient, master)
util.WaitForOffsetSync(t, masterClient, slaveClient, 60*time.Second)

// Namespaces should be replicated after the full sync
token, err := slaveClient.Do(ctx, "NAMESPACE", "GET", "foo").Result()
require.NoError(t, err)
require.EqualValues(t, "bar", token)
}

func TestNamespaceRewrite(t *testing.T) {
password := "pwd"
srv := util.StartServerWithCLIOptions(t, false, map[string]string{
Expand Down
Loading