Write error returned by BlockNotifications() call ins listenLoop method body
to error channel only if it is procided. Otherwise, write debug log message.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Replace block channel's initialization from BlockNotifications method body
to Subscriber's constructor in order to prevent potential writing to nil
channel.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Call handler of the fractional block interval once between base interval
ticks by default. Add option to call handler of fractional block interval
multiple times (N times if fractional interval == BASE_INTERVAL / N).
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Extend Listener with RegisterBlockHandler method. All block handlers are
called on each block read from Subscriber.BlockNotifications channel.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Call SubscribeForNewBlocks in Subscriber's constructor. Provide
BlockNotifications interface method that returns block channel. Write new
blocks to the channel on notification events with BlockEventID type.
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>
Pivot used to shuffle nodes in the CRUSH tree. This is
required argument. We use container ID value to select
container nodes, so `nil` value produces incorrect placements.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Replace static NodeInfo structure with NodeState interface that provides
method to read node information in runtime.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
There is a codecov issue because objects are not placed
in the engine the same way every unit test. Therefore
sometimes there are more coverage, sometimes there are
less. Seeded RNG should solve this issue for engine tests.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Netmap snapshot table caches bootstrap `AddPeer` txs so inner ring
node does not produce redundant approval txs for bootstrapped nodes.
However if node updates states to `Offline`, then such node should
be flagged in snapshot table, so re-bootstrap will actually produce
approval tx.
`ev.PublicKey.String()` returns uncompressed representation of the
node's public key, while snapshot contains compressed values.
Therefore the node was not flagged and re-bootstrap tx was not
approved by inner ring nodes.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Add HealthStatus method to HealthChecker interface that should return
current health status of the node application.
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>
Add NetmapSnapshot rpc to ControlService protobuf definition. Recompile
proto files. Add required method to server structure.
Signed-off-by: Leonard Lyubich <leonard@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>
Add binary salt field to RangeHashPrm struct. Implement field setter. Set
salt from the request in v2 service.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
We can't use enum values from SDK library directly, they can be
different from API specification. Therefore we need to convert them
into protocol level format.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
In previous implementation of eACL service v2 the response X-headers were
validated at the stage of re-checking eACL. This provoked a mismatch of
records in the eACL table with requests. Fix this behavior by checking the
headers from the request, not the response.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Forward request X-headers to client calls during internal processing of
Object operations on the node.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Since PoR audit check uses object.Head with large TTL values to
make real proof of retrievability, we may use random remote
nodes.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Audit task manager should not discard tasks if all workers are
busy, therefore pools should not be non-blocking.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
PDP audit check is not quite working with very small objects, so
we try to build coverage with bigger objects.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
This function converts 2-dimension array of container nodes
into single dimension array. Useful when we need to iterate
over whole container.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Add numeric return from TaskManager.Reset method that shows the number of
canceled tasks. This values will be used for assessment of the progress of
the audit.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Define interface of the container communicator which methods are going to be
used in audit checks. Make innerring Server to implement this interface.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Implement Reset method on audit task manager that cleans task queue.
Extended TaskManager interface with Reset method on IR side. Call Reset
method in audit processor before new audit start.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
In previous implementation parent object header finalized twice in size
limiter + formatter. On the one hand, this added redundant action, on the
other hand, it could provoke a difference in the headers of the linking and
the last part. Change formatter to finalize parent header if it does not
container the signature. Change size limiter to reuse parent header after
last child finalization in linking child.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
In previous implementation DB.Containers method could return an error about
invalid container ID string format. This could happen if some of top-level
buckets had name w/o "_" substring.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
In previous implementation eACL validator didn't take into account container
and object ID fields of request bodies.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
Every unknown error must not decrease shortage counter and must not
exclude faulty node from the node list, because this list will be used
later for replication.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
Deadlock occurs when `getActivate` function opens new blobovnicza and that
invokes evict in LRU cache of open blobovniczas. `getActivate` makes
`activeMtx.Lock()` and then cache evict makes `activeMtx.RLock()` and deadlock
happens.
Fix contains two steps:
- add separate mutex to open blobovniczas (1),
- split single Lock outside of `updateAndGet` (2).
As for the (1) `bbolt.Open()` locks when it tries to open the same file from
two threads. So separate mutex will prevent that.
As for the (2) `updateAndGet` function contains from two parts. At first it
checks if required blobovnicza is ready and it returns it. In this case we can
use the simple RLock. But then there is an option when we should open new
blobovnicza and update map of active blobovniczas.
In this case we call `openBlobovnicza` without activeMtx lock. Cache evict
happens there and it won't cause deadlock.
Then we lock activeMtx to update the map of active blobovniczas. Concurrency can
happen there. However `openBlobovnicza` will not open the same blobovnicza twice,
so we can make one more check if opened blobovnicza was activated while thread was
locked in activeMtx. If so, then return active blobovnicza, else finish activation.
Signed-off-by: Alex Vanin <alexey@nspcc.ru>
In previous implementation Blobovnicza could incorrectly initialize
dimensional buckets: if SmallSizeLimit = 2 ^ X + Y && Y < 2 ^ X, then
largest dimensional bucket was [2 ^ (X - 1) : 2 ^ X]. This was caused by an
incorrect condition for stopping the iterator along the dimensional
boundaries.
Signed-off-by: Leonard Lyubich <leonard@nspcc.ru>
