Prevent some serialization errors that occur due to serialization
and deserialization of `ArrayList$SubList` found inside the
`SessionState` data structures.
To prevent this, an explicit `ArrayList` is used rather than a `List`.
Overload the `List` operator functions so that `+` returns an
`ArrayList` instead of a `List`.
Create `toArrayList` for a few conversions.
If a flow fails outside of its normal error processing code path it will
end up in `FlowDefaultUncaughtExceptionHandler`.
This handler will put the flow into overnight observation if possible.
This is done in-memory and the database.
Even with this being done, the fiber itself has blown up and therefore
does not manage to get to `SMM.removeFlow` which is where
`SMM.decrementLiveFibers` is called. For example, a flow that errored
will hit this code eventually. This code is also hit when a flow is
suspended and a shutdown event is sent to it.
The `liveFibers` latch blocks the SMM from shutting down until all flows
have finished or processed shutdown events.
The changes described below resolve this problem.
Any flow that goes to the `FlowDefaultUncaughtExceptionHandler` will be
put marked as dead (`StateMachineState.isDead`). Highlighting that the
flow cannot continue to process events normally as it has broken out
of its event loop
Retrying and shutdown are done manually rather than injecting events
into the flow fiber's queue, because it can't execute its event loop.
Killing a dead flow executes an altered version of
`retryFlowFromSafePoint`. It does this so it can delete the checkpoint
and then continue using the checkpoint it just deleted to run the
kill flow transition on a new fiber.
If a killed flow reaches the `FlowDefaultUncaughtExceptionHandler` it
will be forcibly killed via `killFlowForcibly` which deletes the
checkpoint/or updates it to KILLED and then calls `removeFlow` to bypass
any event processing. This means that a flow that was dead and was killed
will be terminated manually if it reaches the handler again. The same is
true for flows that were not dead before but reached the handler after
being killed.
Also, `FlowCreator.createFlowFromCheckpoint` now retains the `isKilled`
state of the previous fiber's state.
* CORDA-4045 Disable flaky test (#6792)
Disable `Restart does not set senderUUID` as it is unstable.
* INFRA-965: forward merge from 4.5 (2020-11-15) (#6806)
* EG-4168 Updating contributors.md list for OS 4.5 release branch (#6784)
* INFRA-965: forward merge from 4.4 (2020-11-14) (#6805)
* EG-4168 Updating contributors.md list for OS 4.4 release branch (#6785)
* INFRA-965: forward merge from 4.3 (2020-11-13) (#6803)
* INFRA-965: Jenkins/NexusIQ integration should target patches (#6802)
as well as major/minor releases
* Updated nightly builds to match changes for release branch builds
* Updated JDK11 builds to match changes for release branch builds
Co-authored-by: ivanterziev <61829352+ivanterziev-r3@users.noreply.github.com>
* Backporting updates for check-pr-title workflow
Co-authored-by: ivanterziev <61829352+ivanterziev-r3@users.noreply.github.com>
Co-authored-by: Ross Nicoll <ross.nicoll@r3.com>
Co-authored-by: ivanterziev <61829352+ivanterziev-r3@users.noreply.github.com>
* wip
* wip
* wip (need to review IEE comments)
* wip
* wip
* Small refactoring, fixed network-verifier's TestNotaryFlow
* Added command line option to explicitly enable hash agility support
* wip-do-not-push
* wip
* wip
* wip
* aligned merkletree/transaction hash algorithms
* wip
* Added mixed algorithm support for nodes vs leaves and corrected mixed algorithm tests
* moved global computeNonce and componentHash to DigestService
* added comment for failing test to fix
* wip
* Minor cleanups, added deprecated componentHash/computeNonce
* restored exploratory changes to failing SignedTransaction test
* cleaned up and minor rafactoring
* Fixed some tests with hardcoded hash algorithm
* some changes and cleanups following code review
* WIP commit before large change
* WIP Fixed 3 tests
* WIP removed direct references to randomSHA256() and sha256()
* Updated/added liquibase migrations to support larger hash algorithms
* Reviewed, cleanups, comments, fixes
* removing direct references to sha256()
* WIP verifying obligations test errors
* reviewing obligation/attachment issues with sha3_256
* Full review before PR - intermediate commits
* Reviewed and cleaned up
* Futher cleanup
* Fixed partial tree backward compatible json and cleanups
* all tests passing
* Removed couple of unused imports
* Reworked global componentHash function to avoid deprecated warnings
* replaced SHA3s with some alternate SHA2s
* Removed SHA3-256 and SHA3-512 references
* fixed some tests using non ubiquitous hash algorithms
* Fixed ABI compatibility (not for TransactionBuilder)
* Fixed ABI compatibility to TransactionBuilder
* couple of fixes
* fixed DigestService's randomHash
* Removed constructor with loosely typed args for private constructor of LedgerTransaction class (API removal)
* re-introduced LedgerTransaction deprecated ctor for deserialization
* Add possibility to load CustomMessageDigest bypassing JCA (#6798)
* Change api-current for DigestAlgorithm
* disable flaky tests
* addressed liquibase migration script versions
* Removed TODOs and cleanups
* relaxed privacy salt validation
* Fixed privacy salt test to comply with relaxed validation
* detekt and privacySalt validation
* diff cleanup
* diff cleanup
* removed unused import
* removed PrivacySalt's validateFor method and references
* removed invalid character
Co-authored-by: Denis Rekalov <denis.rekalov@r3.com>
* wip
* wip
* wip (need to review IEE comments)
* wip
* wip
* Small refactoring, fixed network-verifier's TestNotaryFlow
* Added command line option to explicitly enable hash agility support
* wip-do-not-push
* wip
* wip
* wip
* aligned merkletree/transaction hash algorithms
* wip
* Added mixed algorithm support for nodes vs leaves and corrected mixed algorithm tests
* moved global computeNonce and componentHash to DigestService
* added comment for failing test to fix
* wip
* Minor cleanups, added deprecated componentHash/computeNonce
* restored exploratory changes to failing SignedTransaction test
* cleaned up and minor rafactoring
* Fixed some tests with hardcoded hash algorithm
* some changes and cleanups following code review
* WIP commit before large change
* WIP Fixed 3 tests
* WIP removed direct references to randomSHA256() and sha256()
* Updated/added liquibase migrations to support larger hash algorithms
* Reviewed, cleanups, comments, fixes
* removing direct references to sha256()
* WIP verifying obligations test errors
* reviewing obligation/attachment issues with sha3_256
* Full review before PR - intermediate commits
* Reviewed and cleaned up
* Futher cleanup
* Fixed partial tree backward compatible json and cleanups
* all tests passing
* Removed couple of unused imports
* Reworked global componentHash function to avoid deprecated warnings
* replaced SHA3s with some alternate SHA2s
* Removed SHA3-256 and SHA3-512 references
* fixed some tests using non ubiquitous hash algorithms
* Fixed ABI compatibility (not for TransactionBuilder)
* Fixed ABI compatibility to TransactionBuilder
* couple of fixes
* fixed DigestService's randomHash
* Removed constructor with loosely typed args for private constructor of LedgerTransaction class (API removal)
* re-introduced LedgerTransaction deprecated ctor for deserialization
* Add possibility to load CustomMessageDigest bypassing JCA (#6798)
* Change api-current for DigestAlgorithm
* disable flaky tests
Co-authored-by: Denis Rekalov <denis.rekalov@r3.com>
Direct SQL was causing issues with the saving of instants not
maintaining their timezones properly.
Changing to parameterised queries fixed this issue.
Also changed other queries to do the same, since it is good practice to
query in this way.
Change assertion in `Restart does not set senderUUID` to verify a single message has a sender UUID set, rather than the last to be recorded as sent has no sender UUID.
* CORDA-3845: Update BC, log4j, slf4j (#6699)
* CORDA-3845: Update BC to 1.64
* CORDA-3845: Upgraded log4j to 2.12.1
* We can remove the use of Manifests from the logging package so that when _it_ logs it doesn't error on the fact the stream was already closed by the default Java logger.
* Remove the logging package as a plugin
* latest BC version
* Remove old test
* Fix some rebased changes to log file handling
* Update slf4j too
Co-authored-by: Ryan Fowler <fowlerrr@users.noreply.github.com>
Co-authored-by: Adel El-Beik <adel.el-beik@r3.com>
* CORDA-4034 Reduce forkEvery to 15 to attempt to mitigate memory leak.
* ENT-5679 Disable test which triggers OOM
* Run tests on two Jenkins agents
* Fixed processing JUnit test results by Allure
* Add timeouts to VaultObserverExceptionTest
* Revert "CORDA-3845: Update BC, log4j, slf4j (#6699)" to eliminate introduced memory leaks
Co-authored-by: Waldemar Zurowski <waldemar.zurowski@r3.com>
Do not let a user reattach to a flow started by another user.
Reattaching to a flow using startFlowWithClientId for a flow not
started by the current user throws a PermissionException
Reattaching to a flow using reattachFlowWithClientId for a flow not
started by the current user returns null.
finishedFlowsWithClientIds does not return flows started by other
users.
Normal rpc permissions around startFlowWithClientId and
startFlowDynamicWithClientId has also been added.
To allow admins to remove client ids as well as be able to see all the
client ids on the node, admin versions have been added that bypass the
user restrictions. These can be permitted via rpc to only provide
their usage to admins.
* Do not use Security.addProvider(BouncyCastleProvider()) in tests, to avoid disruptions of other tests.
* Forcibly register security providers before starting Jimfs, to resolve a sequencing problem where Jimfs triggers loading of the SFTP filesystem provider, which in turn registers the standard BouncyCastle provider rather than the patched version Corda needs.
* INFRA-683 Move Corda OS release branch builds to serial (#6703)
Co-authored-by: Waldemar Zurowski <waldemar.zurowski@r3.com>
Co-authored-by: Denis Rekalov <denis.rekalov@r3.com>
Co-authored-by: Waldemar Zurowski <waldemar.zurowski@r3.com>
* CORDA-3845: Update BC to 1.64
* CORDA-3845: Upgraded log4j to 2.12.1
* We can remove the use of Manifests from the logging package so that when _it_ logs it doesn't error on the fact the stream was already closed by the default Java logger.
* Remove the logging package as a plugin
* latest BC version
* Remove old test
* Fix some rebased changes to log file handling
* Update slf4j too
Co-authored-by: Ryan Fowler <fowlerrr@users.noreply.github.com>
Co-authored-by: Adel El-Beik <adel.el-beik@r3.com>
* Make existing client id flows re-attachable via rpc 'startFlow' when flows draining mode is enabled
* Fix detekt issue
* Remove unneeded/ unreached waiting on flow's return future
Forcibly register security providers before starting Jimfs, to resolve a sequencing problem where Jimfs triggers loading of the SFTP filesystem provider, which in turn registers the standard BouncyCastle provider rather than the patched version Corda needs.
Do not use Security.addProvider(BouncyCastleProvider()) in tests, to avoid disruptions of other tests.
Co-authored-by: Denis Rekalov <denis.rekalov@r3.com>
If a flow is started with a client id, do not delete it from the
database. Instead, set the status to `KILLED` and store a
`KilledFlowException` in the database.
Keeping it around allows it to be reattached to using the same
mechanisms as active, completed or failed flows. Furthermore, without
this change it could be possible for a flow to be triggered again if
killed while the reconnecting rpc client is being used.
If there is no client id related to the flow, then kill flow keeps its
original behaviour of deleting all traces of the flow.
Flows cannot be killed if they are `COMPLETED`, `FAILED` or `KILLED`
already.
Logs have been added if requests to kill flows in these statuses are
made.
Do not update the status + persist the exception if the flow was killed
during flow initialisation (before persisting its first checkpoint).
Remove the client id mapping if the flow was killed and did not persist
its original checkpoint.
* ENT-5666 Extract shutdown events when retrying a flow
When a flow is retrying, only a select set of events are transferred
over to the new fiber. Shutdown events were not included in this set.
This meant that if a flow retries when an `Event.SoftShutdown` is in its
queue, it will never process it. This causes the node to hang, as the
node awaits `liveFibers`, which never reaches 0 because the shutdown
event is never processed.
To resolve this, `Event.SoftShutdown` is added to the set of events to
extract and reschedule.
* ENT-5666 Don't schedule extra shutdown event
When a flow is stopped by an `Event.SoftShutdown` it will eventually
reschedule another shutdown even when it reaches `SMM.removeFlow`. It
won't actually be processed because the flow returns an abort
continuation. But, it does look odd.
Therefore, it now does nothing instead since that is what it was
implicitly doing.
Created a database snapshot of a clean Corda OS 4.5.1 database, which can now be used for testing by both the node driver and mock network.
The MockNetwork was changed from using an in memory database to using an on disk database, and makes use of the snapshot to speed up setup times.
The Node Driver was changed from defaulting to an in-memory database to defaulting to an on-disk database. Tests that do not specify the type of database to use will thus use an on-disk database. Tests that opt in for an in-memory database will continue to use an in-memory database as before.
The database snapshots are copied to the node directory inside the build folder, therefore, they should be cleaned up after a build.
Co-authored-by: Ross Nicoll <ross.nicoll@r3.com>
* EG-3458 - Missing onError implementation message logged in the node log file with ERROR level - the changes made on top of 4.6 branch
* EG-3458 - Reducing the number of logs by only logging on first consecutive error. Retry without completing the observable
* EG-3458 - Refactor the overly complex method to smaller functions
* EG-3458 - Reducing the number of functions in the class
* INFRA-424 linux1 jenkinsfile
* INFRA-424 full run
* INFRA-424 bigger heap size
* Upgraded DJVM to handle BC - latest version of BC is a multirelease JAR.
When reading JKS keystore if a BC EdDSAPrivateKey is returned then swap for a net.i2p EdDSA private key.
* Temporary downgrade of BC
* Removed the BC EdDSA conversion
* INFRA-424 bigger heap size
* Upgrading Quasar to handle openJ9 different fields.
* INFRA-424: Handle lack of SUPPRESSED_SENTINEL in openj9.
* INFRA-424: If BCEdDSA public or private key is generated convert to net.i2p EdDSA form.
* INFRA-424 bigger heap size
* INFRA-424: On openJ9 only getting upto milli resolution.
* INFRA-424: Handle keystore returning a BCEdDSAPrivateKey.
* INFRA-424: Disable test on JDK11, as it requires the custom cordapp to generate JDK8 contract code, which we now check for.
* INFRA-424: Truncated time test to resolution of millis for openj9.
* INFRA-424 disabling log intensive tests until a fix is developed
* INFRA-424 one more test disabled
* INFRA-424: Disabled a couple of tests failing on openj9.
* INFRA-424: Disabling failing openj9 tests.
* INFRA-424: Disabling test failing on openj9.
* INFRA-424: Ignoring another flaky sleep test on openj9.
* INFRA-424 run integrationTests
* INFRA-424 set timeout to 4 hours
* INFRA-424: Cope with exception message from openj9.
* INFRA-424: Handle the coloured text characters openj9 adds.
* INFRA-424: Disabling test as it is generating JDK11 contract code under JDK11. Currently on JDK8 contract code allowed.
* INFRA-424: Commenting test out for openj9. Output of the processs thats read by the test is sometimes garbled.
* INFRA-424 switching to smoke tests
* INFRA-424 switching to slow integration tests
* INFRA-424 full run
* INFRA-424 moving jenkinsfile
* INFRA-424 removing references
* INFRA-424: Created common IS_OPENJ9 func for ignoring tests.
Co-authored-by: Schife <razvan.codreanu@r3.com>
* CORDA-4003: Now cope with file: prefix not being in class path element.
* CORDA-4003: Switched to new URL type filter.
* CORDA-4003: Switched to a URL comparison. In the string comparison the scheme was removed in latest version of classgraph.
* CORDA-4003: Moved to latest version of classgraph that has support for + in filenames.
* CORDA-4003: Switched to accept version of the deprecated classgraph methods.
* CORDA-3995 Redeliver external events in number of suspends differs
When retrying a flow, only redeliver external events held in a flow's
pending deduplication handlers if there is a difference in the
`numberOfSuspends` on the `currentState`'s checkpoint or the checkpoint
in the database.
If the checkpoint committed, but the flow retried, then the external
events would have been persisted to the database as part of the same
transaction. Therefore there is no need to replay them, as they have
already been processed as saved as part of the checkpoint.
This change is only relevant when the checkpoint persists, but the flow
still needs to retry after this occurs (within the same
transition/event).
* CORDA-3995 Redeliver external events in number of commits differs
When retrying a flow, only redeliver external events held in a flow's
pending deduplication handlers if there is a difference in the
`numberOfCommits` on the `currentState`'s checkpoint or the checkpoint
in the database.
If the checkpoint committed, but the flow retried, then the external
events would have been persisted to the database as part of the same
transaction. Therefore there is no need to replay them, as they have
already been processed as saved as part of the checkpoint.
This change is only relevant when the checkpoint persists, but the flow
still needs to retry after this occurs (within the same
transition/event).
* CORDA-3995 Redeliver external events if number of commits differs
When retrying a flow, only redeliver external events held in a flow's
pending deduplication handlers if there is a difference in the
`currentState`'s `numberOfCommits` or the `numberOfCommits`
the checkpoint has recorded in the database.
If the checkpoint committed, but the flow retried, then the external
events would have been persisted to the database as part of the same
transaction. Therefore there is no need to replay them, as they have
already been processed as saved as part of the checkpoint.
This change is only relevant when the checkpoint persists, but the flow
still needs to retry after this occurs (within the same
transition/event).
* Add @Suspendable to a test flow.
I am surprised this worked at all.
* Fix a few minor things based on review.
Co-authored-by: Will Vigor <william.vigor@r3.com>
We should check that PAUSED Checkpoints can be deserialised on node
startup as we do for RUNNABLE checkpoints. Otherwise a user might
get into trouble if they update the CorDapp.
We should not overwrite the stack trace of local errors thrown by
`FlowContinuation.Throw` as it hides the real cause of the error.
Exceptions received from peer nodes are still overwritten.
* CORDA-3960: Port MultiRPCClient to OS
* CORDA-3960: Carefully restore serialisation environment in `MultiRpcClientTest` to eliminate side effects on other tests
* CORDA-3960: Move ThreadDumpUtils.kt to `core-utils`
Flows that were started with a client id would hang because it would
retrieve the existing flow's future and wait for it to finish. But,
because the flow has failed its flow init and not saved its initial
checkpoint, it is relying on `startFlow` to start the flow again (by
redelivering the start flow external event).
`FlowWithClientIdStatus` now holds the flow id that it is related to.
This is then checked in `startFlow`. If a matching client id is found
for a flow start, it then checks the flow id as well. If the flow id
matches, then it lets the `startFlow` call continue, allowing it to
actually start the flow again (how a flow without a client id would
retry in this situation).
Previously we were just throwing this away when pausing, meaning
updates would not be passed back to the user.
The progress tracker is now maintained in the `NonResidentFlow`
allowing it to be reused in the flow when it is retried.
* ENT-5672 Update database query to get paused flows which have previously been hospitalised
* NOTICK Remove unneeded check if a database exception was removed when switching a flow to RUNNABLE since we were to remove it anyway
Always attempt to load a checkpoint from the database when a flow
retries.
This is to prevent transient errors where the checkpoint is committed to
the database but throws an error back to the node. When the node tries
to retry in this scenario, `isAnyCheckpointPersisted` is false, meaning
that it will try to insert when it tries to save its initial checkpoint
again.
By loading from the existing checkpoint, even though it doesn't
really use it because it is `Unstarted`, the flag gets put into the
right state and will update rather than insert later on.
The issue with the test was that the environment variable are kept as a static member so it passed if it was the first one to run, but failed if another test runs the config beforehand.
Terminate sessions that need to be removed instantly in whatever transition is currently executing, rather than scheduling another event and doing so at a later time.
To do this, update the transition being created in `TopLevelTransition` to remove the sessions and append the `RemoveSessionBindings` action to it.
This achieves the same outcome as the original code but does so with 1 less transition. Doing this also removes the race condition that can occur where another external event is added to the flow's event queue before the terminate event could be added.
* CORDA-3986 Increase sleep in `FlowSessionCloseTest`
A sleep duration needed to be increased to ensure that an end session
message has time to be processed by the other node.
Locks do not fully fix this because some internal processing needs to be
completed that can't be waited for using a lock. Therefore the sleep
time was increased generously.
* removed dependency from tools.jar
I removed the log line in /node/src/main/kotlin/net/corda/node/internal/NodeStartup.kt because I felt it was not so important
and I modified the checkpoint agent detection simply using a static field (I tested both with and without the checkpoint agent running and detection works correctly)
* move method to node-api to address review comments
Co-authored-by: Walter Oggioni <walter.oggioni@r3.com>
* CORDA-3954: Added step to run database migration scripts during initial node registration with -s / --skip-schema-creation option (default to false) to prevent migration.
* CORDA-3954: Applied code convention to if statement
* CORDA-3954: Marked NodeCmdLineOptions' -s/--skip-schema-creation as deprecated and hidden in line with --initial-registration
Disable `SignatureConstraintMigrationFromHashConstraintsTests.HashConstraint cannot be migrated to SignatureConstraint if a HashConstraint is specified for one state and another uses an AutomaticPlaceholderConstraint()` as it frequently appears in reports about Gradle process failures, to try isolating the actual cause.
* fix suggestion and tests
* detekt suppress
* making sure the forced join works with IndirectStatePersistable and removing unnecessary joinPredicates from parse with sorting
* remove joinPredicates and add tests
* rename sorting
* revert deleting joinPredicates and modify the force join to use `OR` instead of `AND`
* add system property switch
* CORDA-3657 Extract information from state machine
`FlowReadOperations` interface provides functions that extract
information about flows from the state machine manager.
`FlowOperator` implements this interface (along with another currenly
empty interface).
* CORDA-3657 Rename function and use set
* initial test is passing
* wip
* done tests
* additional tests to cover more FlowIORequest variations
* completed tests
* The quasar.jar should nat have been changed
* Fixed issues reported by detekt
* got rid of sync objects, instead relying on nodes being offline
* Added extra grouping test and minor simplification
* Hospital test must use online node which fails on otherside
* Added additional information required for the ENT
* Added tests to cover SEND FlowIORequests
* using node name constants from the core testing module
* Changed flow operator to the query pattern
* made query fields mutable to simply building query
* fixed detekt issue
* Fixed test which had dependency on the order int the result (failed for windows)
* Fixed recommendations in PR
* Moved WrappedFlowExternalOperation and WrappedFlowExternalAsyncOperation to FlowExternalOperation.kt as per PR comment
* Moved extension to FlowAsyncOperation
* removed unnecessarily brackets
Co-authored-by: LankyDan <danknewton@hotmail.com>
Fix memory leak due to DB not shutting down in FlowFrameworkPersistenceTests.flow restarted just after receiving payload.
Also reduces number of class-wide variables to reduce scope for references being accidentally held between runs.
`KillFlowTest` is failing quite often. This is probably due to issues in ordering when taking and releasing locks. By using `CountDownLatch` in places instead of `Semaphore`s should reduce the likelihood of tests failing.
If an error occurs when creating a transition (a.k.a anything inside of
`TopLevelTransition`) then resume the flow with the error that occurred.
This is needed, because the current code is swallowing all errors thrown
at this point and causing the flow to hang.
This change will allow better debugging of errors since the real error
will be thrown back to the flow and will get handled and logged by the
normal error code path.
Extra logging has been added to `processEventsUntilFlowIsResumed`, just
in case an exception gets thrown out of the normal code path. We do not
want this exception to be swallowed as it can make it impossible to
debug the original error.
Save the exception for flows that fail during session init when they are
kept for observation.
Change the exception tidy up logic to only update the flow's status if
the exception was removed.
Add `CordaRPCOps.reattachFlowWithClientId` to allow clients to reattach
to an existing flow by only providing a client id. This behaviour is the
same as calling `startFlowDynamicWithClientId` for an existing
`clientId`. Where it differs is `reattachFlowWithClientId` will return
`null` if there is no flow running or finished on the node with the same
client id.
Return `null` if record deleted from race-condition
Update the compatible flag in the DB if the flowstate cannot be deserialised.
The most common cause of this problem is if a CorDapp has been upgraded
without draining flows from the node.
`RUNNABLE` and `HOSPITALISED` flows are restored on node startup so
the flag is set for these then. The flag can also be set when a flow
retries for some reason (see retryFlowFromSafePoint) in this case the
problem has been caused by another reason.
Added a newpause event to the statemachine which returns an Abort
continuation and causes the flow to be moved into the Paused flow Map.
Flows can receive session messages whilst paused.
Add a lock to `StateMachineState`, allowing every flow to lock
themselves when performing a transition or when an external thread (such
as `killFlow`) tries to interact with a flow from occurring at the same
time.
Doing this prevents race-conditions where the external threads mutate
the database or the flow's state causing an in-flight transition to
fail.
A `Semaphore` is used to acquire and release the lock. A `ReentrantLock`
is not used as it is possible for a flow to suspend while locked, and
resume on a different thread. This causes a `ReentrantLock` to fail when
releasing the lock because the thread doing so is not the thread holding
the lock. `Semaphore`s can be used across threads, therefore bypassing
this issue.
The lock is copied across when a flow is retried. This is to prevent
another thread from interacting with a flow just after it has been
retried. Without copying the lock, the external thread would acquire the
old lock and execute, while the fiber thread acquires the new lock and
also executes.
* Remove use of Thread.sleep() FROM FlowReloadAfterCheckpointTest, instead relying on CountdownLatch to wait until the target number has been hit or a timeout occurs, so the thread can continue as soon as the target is hit.
* Replace use of hashmaps to a concurrent queue, to mitigate risk of complex threading issues.
