Application can be shut down often, it is undesirable to send status changes
on a transaction every time. Control API should be used before shutdown for
this purpose.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Make work scope and expansion rate the app-side constants since Policer
independently regulates the amount of work performed.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
There is no need to use synchronous execution of notification handlers. Also
there is no understanding of how to assess the need to change the size of
the pools.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Calls to contracts by storage nodes do not lead to the accumulation of
multisignatures in the contract memory, so the call cost can always be
accurately calculated in advance without additional fee.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Replace dial timeout of API client connection from `grpc` section to
`apiclient` since this value is semantically decoupled from the transport
protocol.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Rename `util/profiler` package to `httputil` and refactor it:
* simplify utility HTTP server;
* make more generic server's parameters in order to remove `viper.Viper`
dependency;
* use single constructor for creating the pprof and prometheus servers;
* replace `enabled` config value with empty-check of the network address.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Establish client connection with main chain node on storage node startup.
Client is configured simlarly to morph client.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Writing application version in each log message seems pretty redundant.
Remove global `app_version` field from Node/IR loggers. Write version, build
time and debug flag once on start.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Writing application name in each log message seems pretty redundant.
Remove no longer used `misc.NodeName` and `misc.InnerRingName` constants.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Replace `viper.Viper` parameter with `Prm` structure. Currently only logging
level can be parameterized through string setter.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Remove setting of `app_name` and `app_version` fields in `NewLogger`
constructor. Set these fields in Node and IR application in already
constructed log and remove them from viper .
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Change logger's encoding to `console`, time encoding to `ISO8601TimeEncoder`
and leave all other options as they are in `zap.NewProductionConfig`.
Remove default values of no longer existing options in node/ir config.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Includes:
- Delete first `ctx` argument in `Write` method.
- Move intermediate Initial trust struct and method
to `calculator` file.
- Change Alpha to 0.1.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Init all wrappers eigenTrust algorithm needed in
`main` packages. Implement `SendIntermediateResult`
GRPC method of `reputationServer`.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Delete reading `alpha` from env var. Cover
retrieving `alpha` behind interface in
intermediate calculator. Add TODO to decide
if it is necessary to receive that param from
global config or not.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Add consumer storage wrapper that implements
`WriterProvider` interface. Change field naming
in daughters storage package.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Move common remoteProvider code to cmd/reputation/common.
Hide WriterProvider initialization behind interface and
add implementation of that interface to local and
intermediate packages in cmd/reputation directory.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Add implementation of Builder interface for
intermediate trusts. Move all code associated
with managers to `common` directory in `cmd`
and `pkg/services/reputation`
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Control service turns storage node to online and offline modes. We can't
reuse `cfgNetmap.reBootstrapEnabled` because it represents configuration
value while control service manages state in runtime.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Eigen trust block timer ticks to start new round of
eigen trust calculations. Every epoch this timer
recalculates duration and starts again.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Add `addContainerAsyncNotificationHandler`
func that passed handler to container's
worker pool similar to how it is done in
reputation handlers.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Add handler closure over worker pool
in the event package.
Add `addNewEpochAsyncNotificationHandler`
function that uses that closure. Pass
the reputation report handler to worker
pool via using that function.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Add worker pools to Netmap and Container
config structures. Add its initialization
that depends on environmental variables(
sync/async; worker pool size).
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
if `shutdown.offline.enabled` config option set to true, then
send netmap.UpdateState(offline) tx at the shutdown. This config
option is not set by default.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Add `Router` to the reputation server. `Router` is
called on every incoming request and inits `Writer`
that sends `Trust`s to the next route point or
handle(logs in that implementation) them if current
node is the end point of the route.
Rename `onlyKeyRemoteServerInfo` struct for container
to separate it from the same implementation of the
same `ServerInfo` interface for reputation.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Implement dependencies for local trust router
in `main` package. Managers are defined as
next node after the current one in HRW sorted
list of current NetMap.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
Register reputation service server that:
* verifies signatures of the requests and signs the responses;
* attaches response meta header to the responses;
* logs body structures of the requests (only as initial implementation, in
future it will be replaces with local trust storage).
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
There is a need to track the results of Object interactions with the remote
peers during node's lifetime. Each successful operation should increment the
number of satisfactory interactions with the remote peer, and the failed
ones, on the contrary, decrement. Collected numbers of satisfactory
interactions are going to be used for calculation of the normalized local
trust values according to original EigenTrust algorithm.
Implement wrapper over local trust storage that calls `Update` method on it
with the result of any object operation. The wrapper provides interface of
the remote client constructor. It is used as a `ClientConstructor` component
in Object service server of the app.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Implement and call `initReputationService` func that constructs local trust
storage and controller, and subscribes the controller on new epoch
notification. Event handler calls `Controller.Report` method to process
collected values.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Implement simple `WriterProvider` building a `Writer` that writes incoming
data to the log. In the future, this action will be replaced by sending the
value to the manager nodes.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Implement iterator provider on wrapper over the local trust storage.
Provided iterator passes normalized local trust values according to original
EigenTrust algorithm description. In the borderline case, normalized values
are set as 1/N, N - number of remote nodes in network map.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Replace usage of `cache.ClientCache` type with interface with similar
signature. This will further allow overloading clients without affecting the
logic of dependent packages.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
ttlNetCache should evict records after TTL duration. However if
data is often accessed and there are no LRU eviction (cache used
with small number of keys), then data will not be evicted ever.
This is a invalid behaviour for mutable data such as eACL.
Solution is to not update access time on every get, so the data
will be guarantee evicted after TTL duration.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
`WithMetrics` can't make nil check without reflection so we
have to explicitly check if metrics enabled outside of
engine constructor.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Because options are not used when client is already in cache
providing them to shared cache is misleading at best.
In the worst case `dial_timeout` is set randomly (because of race
condition) which can lead to one service having `dial_timeout` of
another. Thus we set default client creation options when cache is
created.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Implement LRU cache of eACL tables read from sidechain. Use it as a netmap
storage in neofs-node app. Current cache size is set to 10 (constant).
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Implement TTL cache of eACL tables read from sidechain. Use it as a eACL
storage in neofs-node app. Current cache size is set to 100, TTL is 30s
(constants).
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Implement LRU cache for values read from the network. Implement LRU cache of
containers read from sidechain. Use it as a container storage in neofs-node
app. Current cache size is set to 100, TTL is 30s (constants).
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Replace `eacl.Storage` interface implementation from eACL lib to neofs-node
app package. Remove `eacl.WithMorphClient` option in order to abstract eACL
validator from eACL storage implementation.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Exec `StorageEngine.Inhume` operation on redundant copy callback from Object
Policer with `MarkAsGarbage` parameter.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Re-compile protobuf definition of Control service. Implement required
messages on DropObjects RPC request and response messages. Implement
`DropObjects` method on Control service server of the node. Use
`StorageEngine.Delete` method as a deleted object handler on server.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Implement `NetworkInfo` calls on full stack of Netmap services. Current
epoch is read from node local state, magic number is read via `MagicNumber`
call of morph client.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Add `TOMBSTONE_LIFETIME` configuration value of the node which is measured
in NeoFS epoch and is set to 5 by default.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Split the attributes into those that must be explicitly set in the
configuration, and those that, if absent, will be assigned a default value.
Support this logic in `addWellKnownAttributes` function. If no explicit
attribute is set, the application will panic.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
At startup, update only node info status. Leave all other
attributes from node configuration, so user can update them.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Construct used space Controller on node's app-side. Call Controller.Start on
StartEstimation event from sidechain. Call Controller.Stop on
StopEstimation event from sidechain.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Implement processing of AnnounceUsedSpace RPC from NeoFS API v2 method on
the app-side according to the following algorithm (except common steps for
each RPC):
* check if the request sender is presented in container;
* check route of the request (compose from signatures);
* pass the value to Writer returned by WriterProvider.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Define component on app-side that implements RemoteWriterProvider with next
logic:
* if ServerInfo is nil, call internal dead end provider;
* if address in ServerInfo equals to local node's address, do nothing;
* otherwise create API client instance via client cache and return the
wrapper over it.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Implement wrapper over StorageEngine on app-side that iterates over the
values returned by ListContainers+ContainerSize method pair. Component will
be used as the source of local metrics.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Implement wrapper over netmap and container storages on app-side that builds
placement vectors by next steps:
* read container from storage;
* read network map from storage by epoch number;
* build container nodes according to placement policy;
* build placement vectors with epoch-dependent pivot.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Implement wrapper over API client on app-side that passes used space values
to it. Values are buffered on Put call and sent on Close call.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Implement wrapper over Container contract's client on app-side that passes
used space values to it. Uses a fixed public key, whose value will be the
key of the local storage node.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Classifier looks at list of inner ring nodes and container
nodes from current and previous epoch to classify request.
Sometimes these checks might return error.
Consider there is a request from unknown key and container's
placement policy valid for current epoch and invalid for past
epoch. Classifier tries to find if key belongs to container
node from current epoch -- it is not. Then it tries to find if
key belongs to container node from past epoch and it throws
error, because placement policy is invalid for past epoch.
This is a legit case and classifier should ignore such errors
to provide best effort in matching. The only error classifier
should return is an error when request does not contain
public key to classify it.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Implement NodeState interface required by Netmap service. Make a single
point of updating the state of the node (for both Netmap and Control
services). Protect node info structure from data race.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Call Netmap contract on SetNetmapStatus rpc of Control service (AddPeer
method if new status is ONLINE, UpdatePeerState method otherwise).
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Switch health status to:
* STARTING after the initialization of app resources and before bootup;
* READY after bootup and until shutting down;
* SHUTTING_DOWN before exiting the application.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Update node network status via update status by getting the latest network
map and finding the node in it. This step is performed at the start of the
application and upon notification of a new epoch.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Implement methods required for signatures. Receive network map from netmap
storage, convert it to Control service message and return in response.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Implement HealthChecker on node app structure. Set health status to ONLINE
after node boot. Set health status to OFFLINE on shutdown.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Application should support several NEO endpoints so it can
switch between different RPC nodes when they fail. Application
iterates over endpoints in random order so the default list of
endpoints distribute workload kinda uniformly.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
As in #72 storage application should behave the same way at remote
RPC node failures. The simplest way is to restart application. Later
we can reinitialize it without downtime.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Support processing of NetmapEpoch and NetmapLookupDepth X-headers when
processing object read operations. Placement for operations
Get/Head/GetRange/GetRangeHash/Search is built for the epoch specified in
NetmapEpoch X-header (by default latest). Also the specified operations are
processed until success is achieved for network maps from the past up to
NetmapLookupDepth value. Behavior for default values (zero or missing) left
unchanged.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Transfer the state of the node to offline when the application is shut down.
Updating the state is done by calling UpdateState method of Netmap contract.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
With the update of the local storage engine, the headers of virtual objects
are directly given. In this regard, the step with obtaining the the right
child header is removed.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Define networkState structure and implement netmap.State interface on it.
Set epoch number on NewEpoch notification from chain. Provide
WithNetworkState option to Put service constructor.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
These buckets can be used to store blobs and metadata.
They will be removed as enhanced blob storage will be
implemented for neofs-node. To setup storage type
use `storage.object.type` and `storage.meta.type` params.
Available options:
- inmemory (default)
- boltdb
- filesystem
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
GRPC has default message limit of 4MiB. Since every transmitted
neofs message has to be signed, then original message should
be split into transfer fit structures before signature service.
This commit introduce transport payload splitter for object
service pipeline. This splitter works with stream response
for methods:
- object.Get
- object.Range
- object.Search
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Default values for capacity and price will be 0, so storage node
won't be present in placement. Add these attributes if they are
not explicitly set up by user config.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
