frostfs-node/pkg/services/tree/cache.go
Pavel Karpy 479c5a65e1
All checks were successful
ci/woodpecker/pr/pre-commit Pipeline was successful
ci/woodpecker/push/pre-commit Pipeline was successful
[#322] node: Fix tree svc panic
If a connection has not been established earlier, it stores `nil` in LRU
cache. Cache eviction tries to close every connection (even a `nil` one) and
panics but not crash the app because we are using pools.
That ugly bug also leads to a deadlock where `Unlock` is not called via
`defer` func (and that is the way I found it).

Signed-off-by: Pavel Karpy <p.karpy@yadro.com>
2023-05-04 19:48:51 +03:00

110 lines
2.5 KiB
Go

package tree
import (
"context"
"errors"
"fmt"
"strings"
"sync"
"time"
"git.frostfs.info/TrueCloudLab/frostfs-api-go/v2/pkg/tracing"
"git.frostfs.info/TrueCloudLab/frostfs-node/pkg/network"
"github.com/hashicorp/golang-lru/v2/simplelru"
"google.golang.org/grpc"
"google.golang.org/grpc/connectivity"
"google.golang.org/grpc/credentials/insecure"
)
type clientCache struct {
sync.Mutex
simplelru.LRU[string, cacheItem]
}
type cacheItem struct {
cc *grpc.ClientConn
lastTry time.Time
}
const (
defaultClientCacheSize = 10
defaultClientConnectTimeout = time.Second * 2
defaultReconnectInterval = time.Second * 15
)
var errRecentlyFailed = errors.New("client has recently failed")
func (c *clientCache) init() {
l, _ := simplelru.NewLRU[string, cacheItem](defaultClientCacheSize, func(_ string, value cacheItem) {
if conn := value.cc; conn != nil {
_ = conn.Close()
}
})
c.LRU = *l
}
func (c *clientCache) get(ctx context.Context, netmapAddr string) (TreeServiceClient, error) {
c.Lock()
ccInt, ok := c.LRU.Get(netmapAddr)
c.Unlock()
if ok {
item := ccInt
if item.cc == nil {
if d := time.Since(item.lastTry); d < defaultReconnectInterval {
return nil, fmt.Errorf("%w: %s till the next reconnection to %s",
errRecentlyFailed, d, netmapAddr)
}
} else {
if s := item.cc.GetState(); s == connectivity.Idle || s == connectivity.Ready {
return NewTreeServiceClient(item.cc), nil
}
_ = item.cc.Close()
}
}
cc, err := dialTreeService(ctx, netmapAddr)
lastTry := time.Now()
c.Lock()
if err != nil {
c.LRU.Add(netmapAddr, cacheItem{cc: nil, lastTry: lastTry})
} else {
c.LRU.Add(netmapAddr, cacheItem{cc: cc, lastTry: lastTry})
}
c.Unlock()
if err != nil {
return nil, err
}
return NewTreeServiceClient(cc), nil
}
func dialTreeService(ctx context.Context, netmapAddr string) (*grpc.ClientConn, error) {
var netAddr network.Address
if err := netAddr.FromString(netmapAddr); err != nil {
return nil, err
}
opts := []grpc.DialOption{
grpc.WithBlock(),
grpc.WithChainUnaryInterceptor(
tracing.NewGRPCUnaryClientInteceptor(),
),
grpc.WithChainStreamInterceptor(
tracing.NewGRPCStreamClientInterceptor(),
),
}
// FIXME(@fyrchik): ugly hack #1322
if !strings.HasPrefix(netAddr.URIAddr(), "grpcs:") {
opts = append(opts, grpc.WithTransportCredentials(insecure.NewCredentials()))
}
ctx, cancel := context.WithTimeout(ctx, defaultClientConnectTimeout)
cc, err := grpc.DialContext(ctx, netAddr.URIAddr(), opts...)
cancel()
return cc, err
}