[feat] Add configurable deadline for writers (#11822)

This PR adds deadlines per Write() calls, such
that slow drives are timed-out appropriately and
the overall responsiveness for Writes() is always
up to a predefined threshold providing applications
sustained latency even if one of the drives is slow
to respond.
This commit is contained in:
Harshavardhana
2021-03-18 14:09:55 -07:00
committed by GitHub
parent d46c3c07a8
commit 51a8619a79
16 changed files with 191 additions and 90 deletions

View File

@@ -577,15 +577,6 @@ func (sys *IAMSys) Init(ctx context.Context, objAPI ObjectLayer) {
// Hold the lock for migration only.
txnLk := objAPI.NewNSLock(minioMetaBucket, minioConfigPrefix+"/iam.lock")
// Initializing IAM sub-system needs a retry mechanism for
// the following reasons:
// - Read quorum is lost just after the initialization
// of the object layer.
// - Write quorum not met when upgrading configuration
// version is needed, migration is needed etc.
rquorum := InsufficientReadQuorum{}
wquorum := InsufficientWriteQuorum{}
// allocate dynamic timeout once before the loop
iamLockTimeout := newDynamicTimeout(5*time.Second, 3*time.Second)
@@ -620,12 +611,7 @@ func (sys *IAMSys) Init(ctx context.Context, objAPI ObjectLayer) {
// Migrate IAM configuration, if necessary.
if err := sys.doIAMConfigMigration(ctx); err != nil {
txnLk.Unlock()
if errors.Is(err, errDiskNotFound) ||
errors.Is(err, errConfigNotFound) ||
errors.Is(err, context.DeadlineExceeded) ||
errors.As(err, &rquorum) ||
errors.As(err, &wquorum) ||
isErrBucketNotFound(err) {
if configRetriableErrors(err) {
logger.Info("Waiting for all MinIO IAM sub-system to be initialized.. possible cause (%v)", err)
continue
}
@@ -641,12 +627,7 @@ func (sys *IAMSys) Init(ctx context.Context, objAPI ObjectLayer) {
for {
if err := sys.store.loadAll(ctx, sys); err != nil {
if errors.Is(err, errDiskNotFound) ||
errors.Is(err, errConfigNotFound) ||
errors.Is(err, context.DeadlineExceeded) ||
errors.As(err, &rquorum) ||
errors.As(err, &wquorum) ||
isErrBucketNotFound(err) {
if configRetriableErrors(err) {
logger.Info("Waiting for all MinIO IAM sub-system to be initialized.. possible cause (%v)", err)
time.Sleep(time.Duration(r.Float64() * float64(5*time.Second)))
continue