Previously, the depth was restricted because with BFS the amount of
nodes we have in memory blows up exponentially. With DFS is is linear,
so we can process trees of arbitrary depth.
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
Also, try to fetch object header info from the local storage to find as much
object info as possible for the requests which do not assume returning
object header as a response.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
`Degraded` mode can be set by the administrator if needed.
Modifying operations in this mode can lead node into an inconsistent state
because metabase checks such as lock checking are not performed.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Do not check that a node indeed belongs to the container, because the
synchronization will fail in this case anyway.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Do not return backend type from the service for now, because memory
backend is expected to vanish.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
In case node is down or failing for some reason, we can expect `Dial` to
fail. In case we actively try to replicate and `Dial` always takes 2
seconds, replication-related channels quickly become full. That affects
latency of all other write operations.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Before this commit the replication channel was quickly filled under
heavy load. This lead to the continuously increasing latency for all
write operations. Now it looks better.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Also fix a bug with replicator using the multiaddress instead of
<host>:<port> format expected by gRPC library.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Replace `ErrRangeOutOfBounds` error from `pkg/core/object` package with
`ObjectOutOfRange` from `apistatus` package. That error is returned by
storage node's server as NeoFS API statuses.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Do not calculate and do not write homomorphic hash for containers that were
configured to store objects without hash.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Do not use homomorphic hash in storage group for containers that have
`homomorphic_hashing_disabled` set to `true`.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
If the container ID is not nil and not equal to the container ID in the
request, consider bearer token invalid.
See also nspcc-dev/neofs-api#207.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Allocate memory only if a node chosen as the forwarded request receiver
has responded with a successful status.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
After fixing version fields in forwarded requests, a node does not check
statuses since errors are not covered by direct call error checks.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Error checkers now support wrapped errors so there is no need to
explicitly unwrap errors in `Policer`.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Forwarded requests contained zero version in their meta header. It did not
allow responding with API statuses (`v0.0` version considered to be older
than `v2.11`) to the forwarding node and, therefore, did not allow analyzing
responses.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
`auditor` does not need to request SG: processor will fetch that info before
audit context initialization.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
That allows using `ClientCache` for storage group searching before task
context is initialized. Also, that makes it more general purpose.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
After recent changes in NeoFS SDK Go library session tokens aren't
embedded into `container.Container` and `eacl.Table` structures.
Group value, session token and signature in a structure for container
and eACL.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Cache object that are being processed. That prevents concurrent
object handling when there is a few number of objects and object handling
takes more time that the policer needs for starting that object handling one
more time.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
If placement contains two vectors with intersecting nodes it was possible to
send the object to the nodes twice.
Also optimizes requests: do not ask about storing the object twice from the
same node.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
The node does not support asynchronous object replication anymore, so it
does not need to have replicator worker, channel and `AddTask` function.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
In case we have lots of objects in a single container,
`GetContainerNodes` invoked indirectly by a policer can be seen in
pprof.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Node shouldn't perform eACL verification during GET/HEAD request
processing until full object header is received. Otherwise, for some
eACL tables request may be falsely rejected.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Scenario:
* HEAD request of some object
* 1st eACL record allows op for objects with specific user attribute
* 2nd eACL record forbids op by object ID
* node doesn't store the requested object locally
With this scenario node shouldn't deny request.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
It is redundant to process object headers in responses w/o object field
since result will be the same.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Request processing should not be interrupted in case of local storage
failure since error case in normal for relay nodes.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
ACL service should not deny request on local storage failure since in
this case relay nodes won't be able to continue the operation.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
`log.With` is suitable during initialization, but in other places it induces
some overhead, even when branches with logging are not taken.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Core changes:
* avoid package-colliding variable naming
* avoid using pointers to IDs where unnecessary
* avoid using `idSDK` import alias pattern
* use `EncodeToString` for protocol string calculation and `String` for
printing
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Helper function `client.IsErrObjectNotFound` doesn't support error
unwrapping, so we need to do it on caller side.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
`Policer` should pass list of selected candidates into `WithNodes`
method of `replicator.Task`. In previous implementation `processNodes`
method passed an opposite list: failed nodes and/or the local one.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Not all the NeoFS requests must contain OID in their bodies (or must NOT
contain them at all). Do not pass object address in helper functions, pass
CID and OID separately instead.
Also, fixed NPE in the ACL service: updated SDK library brought errors
when working with `Put` and `Search` requests without OID fields.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
In previous implementation `Policer` considered local object copy as
redundant on processing single placement vector.
Make `Policer` to call redundant copy callback after full placement
processing. Also fix 404 error parsing.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Parse all headers beforehand and reject invalid requests.
Another approach would be to remember the error and check
it after `CalculateAction`, which is a bit faster.
The rule of thumb here is "first validate, then use".
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
It is nice to have different paths for different components and also
check that the information returned is different for different shards.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
The service fetches tombstones from the network via object service, every
request is handled in the following order:
1. checks local LRU cache;
2. checks local storage engine;
3. tries to find object in the placement nodes.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
`Degraded` mode is set automatically after error counter is over the
threshold. `ReadOnly` mode can still be set by an administrator.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Add `WithEncryption` option that passes ECDSA key to the persistent session
storage. It uses 32 bytes from marshalled ECDSA key in ASN.1 DER from in
AES-256 algorithm encryption in Galois/Counter Mode.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Move in-memory session storage to the separate directory of `storage`. It is
done for future support of different kind of session storages.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Return `apistatus.ObjectAccessDenied` error on access violation from ACL
service. Write reason in format of the errors from the previous
implementation. These errors are returned by storage node's server as
NeoFS API statuses.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Replace `core/container.ErrNotFound` error returned by `Source.Get`
interface method with `apistatus.ContainerNotFound` status error. This
error is returned by storage node's server as NeoFS API statuses.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Replace `ErrNotFound`/`ErrAlreadyRemoved` error from
`pkg/core/object` package with `ObjectNotFound`/`ObjectAlreadyRemoved`
one from `apistatus` package. These errors are returned by storage
node's server as NeoFS API statuses.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Return `SessionTokenExpired`/`SessionTokenNotFound` error from
`apistatus` package if private session token is expired/missing. These
errors are returned by storage node's server as NeoFS API statuses.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Make all epoch independent in reputation process. Do not reset any timers
related to reputation. Make it possible to finish iteration after the
unexpected `NewEpoch` event.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
In previous implementation `SignService` converted all `error` values to
`INTERNAL` server failure status. That was done for simplification only.
There is a need to transmit status errors as corresponding status
messages.
Make `SignService` to unwrap errors and convert them to status message
during writing to the response. Non-status errors are converted to
`INTERNAL` server failures. Status errors can also be wrapped in the
depths of the executable code, so `SignService` tries to unwrap them.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
After `putsvc.Service` started to support additional container broadcast
of the saved objects there is no more need to perform broadcast of
tombstone object in `deletesvc.Service`.
Make `putsvc.Service` to perform additional broadcast of `TOMBSTONE`
objects. Remove `broadcastTombstone` stage from `deletesvc.execCtx`,
from now it is encapsulated in `saveTombstone` stage. Remove no longer
needed `putsvc.PutInitPrm.WithTraverseOption` method.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
There are several cases when we need to spread the object around the
container after its primary placement (e.g. objects of type TOMBSTONE).
It'd be convenient to support this feature in `putsvc.Service`.
Add additional stage of container broadcast after the object is stored.
This stage is carried out no more than once and does not affect the
outcome of the main PUT operation.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Make all operations that related to `neofs-api-go` library be placed in `v2`
packages. They parse all v2-versioned structs info `neofs-sdk-go`
abstractions and pass them to the corresponding `acl`/`eacl` packages. `v2`
packages are the only packages that do import `neofs-api-go` library. `eacl`
and `acl` provide public functions that only accepts `sdk` structures.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Remove `Object` and `RawObject` types from `pkg/core/object` package.
Use `Object` type from NeoFS SDK Go library everywhere. Avoid using the
deprecated elements.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
- Update `neofs-sdk-go`:
v0.0.0-20211230072947-1fe37df88f80 => v0.0.0-20220113123743-7f3162110659
- Add client interface that duplicates SDK's client behaviour and new
`MultiAddressClient` interface that has method that iterates over wrapped
clients.
- Also start using simple client mode that does not require parsing statuses
outside the SDK library.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Reverse payload overtake triggers direct payload overtake that
sets status and error. We should not override that.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
It is much more convenient to skip source creation.
Also fix some bugs:
1. `cryptoSource.Int63()` now returns number in [0, 1<<63) as required
by `rand.Source` interface.
2. Replace `cryptoSource.Uint63()` with `cryptoSource.Uint64` to allow
generate uint64 numbers directly (see rand.Source64 docs).
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
In previous implementation `Search` method of transport splitter skipped
responses with empty ID list.
Replace while-loop with do-while one in `TransportSplitter.Search`
method implementation in order to send responses with empty result too.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Without sanity check, container service provides successful response,
even though such request will never be approved by Alphabet nodes.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
When application is being terminated, replicator routine
might be on the object picking phase. Storage is terminated
asynchronously, thus `Select()` may return corresponding
error. If we don't process `context.Done()` in this case,
then application freezes on shutdown.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Add `subnet` command which contains all subnet-related commands. Add
sub-commands:
* `create` for creation;
* `remove` for removal;
* `get` for reading;
* `admin` for admin management;
* `client` for client management.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Continues replication executed in separate pool of goroutines,
so there is no need in worker to handle replication tasks
asynchronously.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Upgrade NeoFS API Go library to version with status returns. Make all API
clients to pull out and return errors from failed statuses. Make signature
service to respond with status if client version supports it.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Add `MAINTENANCE` value to `NetmapStatus` enum in Control API. The status is
going to be used to toggle maintenance mode of the storage node.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
All objects in NeoFS must have owner ID. In previous implementation Object
Delete service handler set owner ID from request session token. If removal
was executed w/o a session, object with tombstone was prepared incorrectly.
In order to fix this node should set its own ID and become an owner of the
tombstone object.
Extend `NetworkInfo` interface required by Object.Delete handler with
`LocalNodeID` method which returns `owner.ID` of the local node. Implement
the method on `networkState` component of storage node application which is
updated on each node state change in NeoFS network map. Set owner returned
by `LocalNodeID` call as tombstone object's owner in Delete handler.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
In previous implementation node returns "access denied" on Object.Put with
object with unset owner. Although object owner must be set, its absence
should not be considered as access error. The same applies to sender key.
Check owner ID and public key emptiness only if sticky bit is set.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
The client needs of the Object service are limited and change not often.
Interface changes of the client library should not affect the operation of
various service packages, if they do not change their requirements for
the provided functionality. To localize the use of the base client and
facilitate further support, an auxiliary package is implemented that will
only be used by the Object service.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
There is no point to pass key storage in parameters because
it can be defined on the service level of application.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
`CommonPrm` structure has private key for remote operations.
It obtained in the beginning of request processing. However,
not every operation triggers remote calls. Therefore, key
might not be used. It is important to avoid early key fetching
because `TokenStore` now returns error if session token does not
exist. This is valid case when container nodes receive request with
session token (for ACL pass) and they should process request locally.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
This is invalid operation for storage nodes that receive part of split
object. While object is signed by session token, the message itself
should be signed by the node key.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Some of the pools are initialized during config initialization,
so it isn't possible currently to release them in one place.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Make the implementation of network info source (Netmap V2 service
dependency) to read MillisecondsPerBlock sidechain parameter and NeoFS
network parameters depending on the client version.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
After storage engine started to limit number of PUT operations there is no
need to limited worker pool in Object Put service.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Session token can be present in both object header and
request meta header. They are the same during initial object
placement.
At the object replication, storage node puts object without
any session tokens attached to the request. If container's eACL
denies object.Put for USER role (use bearer to upload), then
replication might fail on objects with session tokens of the
signed by container owner. It is incorrect, so use session
token directly from request meta header.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
There is a need to check if public key in the RPC response matches the
public key of the related storage node declared in network map.
Define `ErrWrongPublicKey` error. Implement RPC response handler's
constructor `AssertKeyResponseCallback` which checks public key. Construct
handler and pass it to client's option `WithResponseInfoHandler`.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
There is a need to have the ability to expand the data needed for client
construction.
Replace `network.AddressGroup` parameter of client cache interfaces with
`client.NodeInfo`.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
In previous implementation Object PUT used single pool of workers for local
and remote ops, but these ops are heterogeneous.
Use remote/local pool for remote/local operations in PUT service. At first
the pools are configured with the same size.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
In previous implementation `distributedTarget` didn't check if next node is
local. This check was performed by the handlers (target initializer and
relay func).
Make `distributedTarget` to calculate node's locality. Pass locality flag to
the handlers.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Forwarding mechanism resends original request. During split object chain traversal,
storage node performs multiple `object.Head` requests on each child. If request
forwarding happens, then `object.Head` returns object ID of the original request.
This produces infinite assembly loop.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Make `errIncompletePut` to be a structure which wraps single client error.
Wrap error of the last client into `errIncompletePut` during placement
execution.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
In previous implementation Object service's handler returned const error in
case of failure (full or partial) of PUT operation. This did not even allow
us to roughly guess what the reason is. Not as a complete solution, but to
alleviate some cases where all nodes in a container return the same error,
it is suggested to return the error of the last server that responded.
Return latest server error from placement loop of `iteratePlacement` method
of `distributedTarget` type.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
In previous implementation sticky bit could disrupt access of container
nodes to replication. According to NeoFS specification sticky bit should not
affect the requests sent by nodes from SYSTEM group.
Add role check to `stickyBitCheck`.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Some software components regulate the way of working with placement arrays
when a local node enters it. In the previous implementation, the locality
criterion was the correspondence between the announced network address
(group) and the address with which the node was configured. However, by
design, network addresses are not unique identifiers of storage nodes in the
system.
Change comparisons by network addresses to comparisons by keys in all
packages with the logic described above. Implement `netmap.AnnouncedKeys`
interface on `cfg` type in the storage node application.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
There is a need to process public keys of the placement result.
Implement `Node.PublicKey` method which returns storage node's key announced
in netmap.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
In previous implementation `placement.Traverser.Next` method returned slice
of `network.AddressGroup` elements. There is a need to process keys of
storage nodes besides network addresses for intra-container communication.
Wrap `network.AddressGroup` in a new type `placement.Node` that summarizes
the storage node information required for communication. Return slice of
`Node` instances from `Traverser.Next` method. Fix compilation breaks in
dependent packages.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
In previous implementation Container service handlers didn't cache the
results of `Get` / `GetEACL` / `List` operations. As a consequence of this,
high load on the service caused neo-go client's connection errors. To avoid
this there is a need to use cache. Object service already uses `Get` and
`GetEACL` caches.
Implement cache of `List` results. Share already implemented cache of Object
service with the Container one. Provide new instance of read-only container
storage (defined as an interface)to morph executor's constructor on which
container service is based. Write operations remained unchanged.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
In previous implementation of Policer's job queue the same list of objects for
processing was selected at each iteration. This was caused by consistent
return of `engine.List` function.
Use `rand.Shuffle` function to compose pseudo-random list of all objects in
order to approximately evenly distribute objects to work.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
