Summary:
It doesn't make sense to tie skipping a member for serialization to a
particular serialization implementation.
Reviewers: mtomic, llugovic
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1743
Summary:
With quite frequent changes of serialization backend, it's getting
really painful updating the C++ implementation. This diff defines the
Symbol class via LCP, so that the C++ serialization is generated instead
of written by hand.
Reviewers: mtomic, llugovic
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1741
Summary:
Instantiation with VertexAccessor was never used, so the template
needlessly complicated the rest of the codebase.
Reviewers: mtomic, llugovic
Reviewed By: mtomic
Subscribers: mferencevic, pullbot
Differential Revision: https://phabricator.memgraph.io/D1736
Summary:
TypeInfo will be needed for upcoming serialization via SLK. This diff
changes the already defined TypeInfo by removing the reliance on
capnp::typeId calls. The struct itself is now in utils.
Hopefully, this shouldn't break our RPC stack due to new ID generation.
Reviewers: mferencevic, mtomic, llugovic
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1735
Summary:
This is the initial work on transferring our serialization code from
Cap'n Proto to SaveLoadKit. The commit contains tests for generated
code, but still requires work on supporting some features. Most notably,
generating and storing type IDs for derived classes so that they can be
loaded from a base pointer. Naturally, loading implementation hasn't
even been started yet.
Reviewers: mtomic, llugovic, mferencevic
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1733
Summary:
This change makes HierarchicalTreeVisitor visit only Cypher related AST
nodes. QueryVisitor can be used to differentiate between various query
types we have. The next step is to either rename HierarchicalTreeVisitor
to something like CypherQueryVisitor, or perhaps extract Clause visiting
from it.
Reviewers: mtomic, llugovic
Reviewed By: llugovic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1710
Summary:
This is the initial step in making the visitors more granular in what
they visit.
Also add ignoring multiple inheritance in LCP and update LCP docs.
Reviewers: mtomic, llugovic
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1704
Summary:
Generate conversion functions from C++ enum to Cap'n Proto and vice
verse. This should reduce the generated code size and hopefully improve
readability.
Use conversion functions for enums in default serialization generation.
`capnp-save-enum` and `capnp-load-enum` should now be used exclusively
to serialize non LCP defined enumerations. They now always require
`enum-values` argument.
Reviewers: mtomic, llugovic
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1696
Summary: this should reduce parsing time for very simple queries.
Reviewers: teon.banek, mferencevic
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1698
Summary:
Blocking transaction has the ability to stop the transaction engine from
starting new transactions (regular or blocking) and to wait all other active
transactions to finish (to become non active, committed or aborted). One thing
that blocking transactions support is defining the parent transaction which
does not need to end in order for the blocking one to start. This is because of
a use case where we start nested transactions.
One could thing we should build indexes inside those blocking transactions. This
is true and I wanted to implement this, but this would require some digging in
the interpreter which I didn't want to do in this change.
Reviewers: mferencevic, vkasljevic, teon.banek
Reviewed By: mferencevic, teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1695
Summary:
`Query` is now an abstract class which has `CypherQuery`,
`ExplainQuery`, `IndexQuery`, `AuthQuery` and `StreamQuery` as derived
classes. Only `CypherQuery` is forwarded to planner and the rest of the
queries are handled directly in the interpreter. This enabled us to
remove auth, explain and stream operators, clean up `Context` class and
remove coupling between `Results` class and plan cache. This should make
it easier to add similar functionality because no logical operator
boilerplate is needed. It should also be easier to separate community
and enterprise features for open source.
Remove Explain logical operator
Separate IndexQuery in AST
Handle index creation in interpreter
Remove CreateIndex operator and ast nodes
Remove plan cache reference from Results
Move auth queries out of operator tree
Remove auth from context
Fix tests, separate stream queries
Remove in_explicit_transaction and streams from context
Reviewers: teon.banek, mferencevic, msantl
Reviewed By: teon.banek, mferencevic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1664
Summary:
The row elements must not be moved, because the following pulls may
still use them.
Reviewers: mtomic, llugovic, mferencevic
Reviewed By: llugovic, mferencevic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1672
Summary:
LabelPropertyIndex now has the ability to enforce unique constraint.
This doesn't lock the tx engine.
Reviewers: teon.banek, mferencevic
Reviewed By: teon.banek
Subscribers: pullbot, vkasljevic, buda
Differential Revision: https://phabricator.memgraph.io/D1660
Summary: Up till now, `AstStorage` also took care of tracking the root of the `Query` and loading of cloning of `Query` nodes would change that root. This felt out of place because sometimes `AstStorage` is used only for storing expressions, and we don't even have an entire query in the storage. This diff removes that feature from `AstStorage`. Now its only functionality is owning AST nodes and assigning unique IDs to them.
Reviewers: teon.banek, llugovic
Reviewed By: teon.banek
Subscribers: mferencevic, pullbot
Differential Revision: https://phabricator.memgraph.io/D1646
Summary:
This diff fixes the storage part of the bug that happen for the
following queries if all nodes end up on the non-master machine.
```
CREATE (a:T{c: True})-[:X{x: 2.5}]->(:A:B)-[:Y]->()-[:Z{r: 1}]->(a)
MATCH (:T{c: True})-[a:X{x: 2.5}]->(node:A:B)-[:Y]->()-[:Z{r: 1}]->() RETURN a AS node, node AS a
```
The current state of the code doesn't work completely, it's still missing the
query execution fix.
Reviewers: teon.banek, ipaljak, vkasljevic
Reviewed By: teon.banek
Subscribers: mferencevic, pullbot
Differential Revision: https://phabricator.memgraph.io/D1629
Summary:
Start removing `is_remote` from `Address`
Remove `GlobalAddress`
Remove `GlobalizedAddress`
Remove bitmasks from `Address`
Remove `is_local` from `Address`
Remove `is_local` from `RecordAccessor`
Remove `worker_id` from `Address`
Remove `worker_id` from `GidGenerator`
Unfriend `IndexRpcServer` from `Storage`
Remove `LocalizedAddressIfPossible`
Make member private
Remove `worker_id` from `Storage`
Copy function to ease removal of distributed logic
Remove `worker_id` from `WriteAheadLog`
Remove `worker_id` from `GraphDb`
Remove `worker_id` from durability
Remove nonexistant function
Remove `gid` from `Address`
Remove usage of `Address`
Remove `Address`
Remove `VertexAddress` and `EdgeAddress`
Fix Id test
Remove `cypher_id` from `VersionList`
Remove `cypher_id` from durability
Remove `cypher_id` member from `VersionList`
Remove `cypher_id` from database
Fix recovery (revert D1142)
Remove unnecessary functions from `GraphDbAccessor`
Revert `InsertEdge` implementation to the way it was in/before D1142
Remove leftover `VertexAddress` from `Edge`
Remove `PostCreateIndex` and `PopulateIndexFromBuildIndex`
Split durability paths into single node and distributed
Fix `TransactionIdFromWalFilename` implementation
Fix tests
Remove `cypher_id` from `snapshooter` and `durability` test
Reviewers: msantl, teon.banek
Reviewed By: msantl
Subscribers: msantl, pullbot
Differential Revision: https://phabricator.memgraph.io/D1647
Summary:
`heaptrack` shows a miniscule decrease in memory usage during query
execution.
Running the below query on the TEDTalk database 100 times gives the
following results:
- number of allocations: from 642647 to 642589
- bytes allocated in total: from 48.79 MiB to 48.78 MiB
```
MATCH (t:Tag)<-[:HasTag]-(n:Talk)
RETURN t.name AS Tag, COUNT(n) AS TalksCount
ORDER BY TalksCount DESC, Tag LIMIT 20;
```
Regarding `TypedValue`'s destructor: we're allowed to manually destruct the
union memebers that we constructed using placement-new. However, it is undefined
behavior to call the destructor after an object's lifetime has ended. Calling
`TypedValue`'s own destructor within its assignment operator counts as ending
its lifetime, which means that the next call to its destructor will invoke
undefined behavior.
See the C++ Draft N4140, clauses 12.4/15 and 3.8/1.3.
Reviewers: teon.banek, mtomic
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1630
Summary:
To clean the working directory after this diff you should execute:
```
rm src/database/counters_rpc_messages.capnp
rm src/database/counters_rpc_messages.hpp
rm src/database/serialization.capnp
rm src/database/serialization.hpp
```
Reviewers: teon.banek, msantl
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1636
Summary: Move explain to Cypher.g4 so MemgraphCypher contains only enterprise features. Also added `EXPLAIN` to keyword list because query `MATCH (explain) RETURN explain` wasn't working.
Reviewers: teon.banek
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1632
Summary:
This should fix the issue of having a write followed by a read during
distributed execution. Any kind of merger of plans should behave like a
Cartesian with regards to planning the following ScanAll.
Reviewers: mtomic, mferencevic
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1624
Summary:
This diff splits single node and distributed storage from each other.
Currently all of the storage code is copied into two directories (one single
node, one distributed). The logic used in the storage implementation isn't
touched, it will be refactored in following diffs.
To clean the working directory after this diff you should execute:
```
rm database/state_delta.capnp
rm database/state_delta.hpp
rm storage/concurrent_id_mapper_rpc_messages.capnp
rm storage/concurrent_id_mapper_rpc_messages.hpp
```
Reviewers: teon.banek, buda, msantl
Reviewed By: teon.banek, msantl
Subscribers: teon.banek, pullbot
Differential Revision: https://phabricator.memgraph.io/D1625
Summary:
TODO:
~~1. Figure out how to propagate exceptions during lambda evaluation to master.~~
~~2. Make some more complicated test cases to see if everything is~~
~~sent over the network properly (lambdas depending on frame, evaluation context).~~
~~3. Support only `GraphView::OLD`.~~
4. [MAYBE] Send only parts of the frame necessary for lambda evaluation.
~~5. Fix EdgeType handling~~
--------------------
Serialize frame and send it in PrepareForExpand RPC
Move Lambda out of ExpandVariable
Send symbol table and filter lambda in CreateBfsSubcursor RPC
Evaluate filter lambda during the expansion
Send evaluation context in CreateBfsSubcursor RPC
Reviewers: teon.banek, msantl
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1600
Summary:
This should allow us to more easily decouple the code which should be
open sourced. Unfortunately, the downside of this approach is that we
cannot rely on virtual calls to dispatch the serialization to correct
type. Another downside is that members need to be publicly accessible
for serialization.
Reviewers: mtomic, msantl
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1596
Summary:
This diff changes the RPC layer to directly return `TResponse` to the user when
issuing a `Call<...>` RPC call. The call throws an exception on failure
(instead of the previous return `nullopt`).
All servers (network, RPC and distributed) are set to have explicit `Shutdown`
methods so that a controlled shutdown can always be performed. The object
destructors now have `CHECK`s to enforce that the `AwaitShutdown` methods were
called.
The distributed memgraph is changed that none of the binaries (master/workers)
crash when there is a communication failure. Instead, the whole cluster starts
a graceful shutdown when a persistent communication error is detected.
Transient errors are allowed during execution. The transaction that errored out
will be aborted on the whole cluster. The cluster state is managed using a new
Heartbeat RPC call.
Reviewers: buda, teon.banek, msantl
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1604
Summary:
Even though we may not need the results of a RPC, we should check that
it completed without error.
Reviewers: mferencevic, mtomic
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1610
Summary:
Knowledge of distributed operators is now removed from PlanPrinter
class. The DistributedOperatorVisitor and PlanPrinter are modified so
that they may be multiple inherited. This is done using virtual
inheritance of HierarchicalLogicalOperatorVisitor. Multiple inheritance
is used to derived a DistributedPlanPrinter which knows how to print
distributed in operators. This removes the dependency of single node
pretty printing on distributed.
Reviewers: msantl, mtomic
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1606
Summary:
Previous version of code was wrong in some cases. Example:
There's an edge e = (a)->(b). Node (a) belongs to worker 1, node (b) belongs to worker 0. Therefore, edge e belongs to worker 1.
When edge e was serialized on worker 0, address of node b was local and it would be globalized, but the worker id of edge (e) would be used, which is wrong.
Reviewers: teon.banek
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1599
Summary:
In a bunch of places `TypedValue` was used where `PropertyValue` should be. A lot of times it was only because `TypedValue` serialization code could be reused for `PropertyValue`, only without providing callbacks for `VERTEX`, `EDGE` and `PATH`. So first I wrote separate serialization code for `PropertyValue` and put it into storage folder. Then I fixed all the places where `TypedValue` was incorrectly used instead of `PropertyValue`. I also disabled implicit `TypedValue` to `PropertyValue` conversion in hopes of preventing misuse in the future.
After that, I wrote code for `VertexAccessor` and `EdgeAccessor` serialization and put it into `storage` folder because it was almost duplicated in distributed BFS and pull produce RPC messages. On the sender side, some subset of records (old or new or both) is serialized, and on the reciever side, records are deserialized and immediately put into transaction cache.
Then I rewrote the `TypedValue` serialization functions (`SaveCapnpTypedValue` and `LoadCapnpTypedValue`) to not take callbacks for `VERTEX`, `EDGE` and `PATH`, but use accessor serialization functions instead. That means that any code that wants to use `TypedValue` serialization must hold a reference to `GraphDbAccessor` and `DataManager`, so that should make clients reconsider if they really want to use `TypedValue` instead of `PropertyValue`.
Reviewers: teon.banek, msantl
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1598
Summary:
This is a simple change which modifies interface of
awesome_memgraph_functions to accept C-style pointer to array with
count. Doing things this way, allows us to easily try out different
allocation schemes for function arguments. In this diff, we are now
using stack allocation of arguments in a plain fixed size array. This is
done when the number of arguments is small. According to heaptrack, this
small change should yield noticeable improvements to heap usage.
Obviously, this doesn't solve the problem of heap allocations inside
TypedValue arguments themselves. These allocations appear when
std::string and std::vector is used inside TypedValue.
Micro benchmarks show that there is some performance improvement,
mostly around the limits of using array vs std::vector. The improvement is
more noticeable with multiple threads, due to primary gain being in avoiding
calls to memory allocation.
Reviewers: mtomic, msantl, mferencevic
Reviewed By: mferencevic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1581
Summary: A quick clean-up of user visible error messages. Tried to make them gramatically correct by capitalizing the first word in the sentence and putting a dot at the end.
Reviewers: teon.banek, buda, ipaljak
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1571
Summary:
Changed GRANT ROLE to SET ROLE. Now it is `SET ROLE FOR user TO ROLE` instead of `GRANT ROLE role TO user`. It makes more sense because our users can only have 1 role.
Changed REVOKE ROLE to CLEAR ROLE. Now it is `CLEAR ROLE FOR user` instead of `REVOKE ROLE role FOR user`. REVOKE ROLE would throw exception if user was not a member of role. CLEAR ROLE clears the role whatever it is. I find that the latter makes more sense combined with SET ROLE.
Changed `SHOW ROLE FOR USER user` to `SHOW ROLE FOR user`.
Changed `SHOW USERS FOR ROLE role` to `SHOW USERS FOR role`.
Reviewers: mferencevic, teon.banek, buda
Reviewed By: mferencevic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1572
Summary:
Visitor pattern's main issue is cyclical dependency between classes that
are visited and the visitor instance itself. We need to decouple this
dependency if we want to open source part of the code, namely
non-distributed part. This decoupling is achieved through the use of
`dynamic_cast` in distributed operators. Hopefully the solution is good
enough and doesn't cause performance issues. An alternative solution is
to build our own custom double dispatch solution, but that will
basically boil down to our implementation of runtime type information
and casts.
Note, this only decouples the distributed operators. If and when we
decide that other operators shouldn't be open sourced, the same
`dynamic_cast` pattern should be applied in them also.
Depends on D1563
Reviewers: mtomic, msantl, buda
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1566
Summary:
This is the first step in separating the implementation of distributed
features in operators. Following steps are:
* decoupling distributed visitors
* injecting distributed details in operator state
* minor cleanup or anything else that was overlooked
Reviewers: mtomic, msantl
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1563
Summary:
The Kafka Python transform functionality uses a Python script to transform
incoming Kafka data into queries and parameters that are executed against the
database. When starting the Python transform script it is started in a
sandboxed environment so that it can't do harm to the host system or the
database.
Reviewers: msantl, teon.banek
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1509
Summary:
Since we switched to Cap'n Proto serialization there's no need for
keeping boost around anymore.
Reviewers: mtomic, mferencevic, buda
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1515
Summary: first step for a cleaner query front end that might be easier to open source
Reviewers: teon.banek
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1511
Summary:
GraphDbAccessor is now constructed only through GraphDb. This allows the
concrete GraphDb to instantiate a concrete GraphDbAccessor. This allows
us to use virtual calls, so that the implementation may be kept
separate. The major downside of doing things this way is heap allocation
of GraphDbAccessor. In case it turns out to be a real performance
issues, another solution with pointer to static implementation may be
used.
InsertVertexIntoRemote is now a non-member function, which reduces
coupling. It made no sense for it to be member function because it used
only the public parts of GraphDbAccessor.
Reviewers: msantl, mtomic, mferencevic
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1504
Summary:
This change, hopefully, simplifies the implementation of different kinds
of GraphDb. The pimpl idiom is now simplified by removing all of the
crazy inheritance. Implementations classes are just plain data stores,
without any methods. The interface classes now have a more flat
hierarchy:
```
GraphDb (pure interface)
|
+----+---------- DistributedGraphDb (pure interface)
| |
Single Node +-----+------+
| |
Master Worker
```
DistributedGraphDb is used as an intermediate interface for all the
things that should work only in distributed. Therefore, virtual calls
for distributed stuff have been removed from GraphDb. Some are exposed
via DistributedGraphDb, other's are only in concrete Master and Worker
classes. The code which relied on those virtual calls has been
refactored to either use DistributedGraphDb, take a pointer to what is
actually needed or use dynamic_cast. Obviously, dynamic_cast is a
temporary solution and should be replaced with another mechanism (e.g.
virtual call, or some other function pointer style).
The cost of the above change is some code duplication in constructors
and destructors of classes. This duplication has a lot of little tweaks
that make it hard to generalize, not to mention that virtual calls do
not work in constructor and destructor. If we really care about
generalizing this, we should think about abandoning RAII in favor of
constructor + Init method.
The next steps for splitting the dependencies that seem logical are:
1) Split GraphDbAccessor implementation, either via inheritance or
passing in an implementation pointer. GraphDbAccessor should then
only be created by a virtual call on GraphDb.
2) Split Interpreter implementation. Besides allowing single node
interpreter to exist without depending on distributed, this will
enable the planner and operators to be correctly separated.
Reviewers: msantl, mferencevic, ipaljak
Reviewed By: msantl
Subscribers: dgleich, pullbot
Differential Revision: https://phabricator.memgraph.io/D1493
Summary: This is a simple thing that enables us to use keywords as symbolic names. A bad thing is that planning will be done for queries which differ only in keyword case (e.g. "MATCH (x) RETURN x" and "match (x) return x"), but that should not be a significant performance impact as queries are done programmaticaly when high throughput is expected and we don't expect letter case change in that case.
Reviewers: buda, mferencevic, teon.banek
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1496
Summary:
First iteration in implementing kafka.
Currently, memgraph streams won't use the transform script provided in the
`CREATE STREAM` query.
There is a manual test that serves a POC purpose which we'll use to fully wire
kafka in memgraph.
Since streams need to download the script, I moved curl init from
telemetry.
Reviewers: teon.banek, mferencevic
Reviewed By: mferencevic
Subscribers: ipaljak, pullbot, buda
Differential Revision: https://phabricator.memgraph.io/D1491
Summary:
This change should preclude the need to specify `:capnp-save` and
`:capnp-load` functions for regularly saved elements of `std::vector<T>`
and `std::optional<T>`. Regular saving in this context means saving
primitive types or compound types which have a `Save(capnp::Builder *)`
method.
Reviewers: mtomic, msantl
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1497
Summary:
Previously, our implementation of the Bolt protocol buffered all results in
memory before sending them out to the client. This implementation immediately
streams the results to the client to avoid any memory allocations. Also, this
implementation splits the interpretation and pulling logic into two.
Reviewers: teon.banek
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1495
Summary:
This change should completely support planning Optional for distributed
execution. Cartesian matching is handled, as well as dependencies
between optional branch and the main input branch.
Unit tests are expanded to cover the planning algorithm.
Reviewers: msantl, mtomic
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1480
Summary:
Session specifics have been move out of the Bolt `executing` state, and
are accessed via pure virtual Session type. Our server is templated on
the session and we are setting the concrete type, so there should be no
virtual call overhead. Abstract Session is used to indicate the
interface, this could have also been templated, but the explicit
interface definition makes it clearer.
Specific session implementation for running Memgraph is now implemented
in memgraph_bolt, which instantiates the concrete session type. This may
not be 100% appropriate place, but Memgraph specific session isn't
needed anywhere else.
Bolt/communication tests now use a dummy session and depend only on
communication, which significantly improves test run times.
All these changes make the communication a library which doesn't depend
on storage nor the database. Only shared connection points, which aren't
part of the base communication library are:
* glue/conversion -- which converts between storage and bolt types, and
* communication/result_stream_faker -- templated, but used in tests and query/repl
Depends on D1453
Reviewers: mferencevic, buda, mtomic, msantl
Reviewed By: mferencevic, mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1456
Summary:
This is the first step in cutting the crazy dependencies of
communication module to the whole database. Includes have been
reorganized and conversion between DecodedValue and other Memgraph types
(TypedValue and PropertyValue) has been extracted to a higher level
component called `communication/conversion`. Encoder, like Decoder, now
relies only on DecodedValue. Hopefully the conversion operations will
not significantly slow down streaming Bolt data.
Additionally, Bolt ID is now wrapped in a class. Our storage model uses
*unsigned* int64, while Bolt expects *signed* int64. The implicit
conversions may lead to encode/decode errors, so the wrapper should
enforce some type safety to prevent such errors.
Reviewers: mferencevic, buda, msantl, mtomic
Reviewed By: mferencevic, mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1453
Summary:
Added test stream functionality. When a stream is configured, it will try to
consume messages from a kafka topic and return them back to the user.
For now, the messages aren't transformed, so it just returns the payload string.
Depends on D1466
Next steps are persisting stream metadata and transforming messages in order to
store them in the graph.
Reviewers: teon.banek, mtomic
Reviewed By: teon.banek
Subscribers: pullbot, buda
Differential Revision: https://phabricator.memgraph.io/D1474
Summary:
Integrated kafka library into memgraph. This version supports all opencypher
features and will only output messages consumed from kafka.
Depends on D1434
Next steps are persisting stream metadata and transforming messages in order to
store them in the graph.
Reviewers: teon.banek, mtomic, mferencevic, buda
Reviewed By: teon.banek
Subscribers: mferencevic, pullbot, buda
Differential Revision: https://phabricator.memgraph.io/D1466
Summary: When doing bfs with given endpoint, we can stop the traversal on first successful pull.
Reviewers: teon.banek, msantl, mculinovic, buda
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1469
Summary:
Added basic functionality for kafka streams. The `CREATE STREAM` clause is a
simplified version from the one mentioned in D1415 so we can start testing
end-to-end sooner.
This diff also includes a bug fix in `lcp.list ` for operators that have no
members.
Reviewers: teon.banek, mtomic, buda
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1434
Summary:
For reasons unknown to man, antlr didn't want to parse calls
to functions whose name is a single hex letter properly. Now it should work.
Reviewers: teon.banek, buda
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1464
Summary:
Added missing string functions.
I also cleaned up error messages a bit in effort to make them uniform.
Reviewers: teon.banek, buda
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1458
Summary:
Hopefully, the mechanism of generating Cartesian is general enough, so
this simple change should work correctly in all cases.
Planner tests have been modified to use a FakeDbAccessor in order to
speed them up and potentially allow extracting planning into a library.
Reviewers: msantl, mtomic, buda
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1431
Summary:
This change should correctly plan Cartesian which have dependent Filter
or Expand operators. Tests have been added for those cases. Other cases
are not yet supported and should throw an exception.
Reviewers: msantl, mtomic, buda
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1426
Summary:
This is the initial step to getting a correct version of distributed
planning of Cartesian operator. Functions and structs have been added
which should collect enough information to correctly order the execution
with regards to dependencies among Cartesian branches. The support
functionality should be the same as was before, but unsupported cases
should now raise an exception instead of leading to undefined behaviour.
Reviewers: msantl, mtomic, buda
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1418
Summary:
Converts the RPC stack to use Cap'n Proto for serialization instead of
boost. There are still some traces of boost in other places in the code,
but most of it is removed. A future diff should cleanup boost for good.
The RPC API is now changed to be more flexible with regards to how
serialize data. This makes the simplest cases a bit more verbose, but
allows complex serialization code to be correctly written instead of
relying on hacks. (For reference, look for the old serialization of
`PullRpc` which had a nasty pointer hacks to inject accessors in
`TypedValue`.)
Since RPC messages were uselessly modeled via inheritance of Message
base class, that class is now removed. Furthermore, that approach
doesn't really work with Cap'n Proto. Instead, each message type is
required to have some type information. This can be automated, so
`define-rpc` has been added to LCP, which hopefully simplifies defining
new RPC request and response messages.
Specify Cap'n Proto schema ID in cmake
This preserves Cap'n Proto generated typeIds across multiple generations
of capnp schemas through LCP. It is imperative that typeId stays the
same to ensure that different compilations of Memgraph may communicate
via RPC in a distributed cluster.
Use CLOS for meta information on C++ types in LCP
Since some structure slots and functions have started to repeat
themselves, it makes sense to model C++ meta information via Common Lisp
Object System.
Depends on D1391
Reviewers: buda, dgleich, mferencevic, mtomic, mculinovic, msantl
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1407
Summary:
Add additional structs and functions for handling C++ meta information.
Add capnp-file and capnp-id arguments to lcp:process-file.
Generate cpp along with hpp and capnp in lcp.
Wrap LogicalOperator base class in lcp:define-class.
Modify logical operators for capnp serialization.
Add query/common.capnp.
Reviewers: mculinovic, buda, mtomic, msantl, ipaljak, dgleich, mferencevic
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1391
Summary:
Command id is necessary in remote produce to identify an ongoing pull
because a transaction can have multiple commands that all belong under
the same plan and tx id.
Reviewers: teon.banek, mtomic, buda
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1386
Summary:
Removing `AS_IS` from GraphView because it doesn't seem like it is necessary for query execution and it also has weird semantics (you might get a mix of old and new records). `Unwind`, `OrderBy` and `PullRemoteOrderBy` now use `OLD` graph view.
Remove AS_IS from GraphView
Fix query_cost_estimator tests
Fix query_expression_evaluator tests
Fix query_plan_match_filter_return tests
Fix query_plan_create_set_remove_delete tests
Fix query_plan_accumulate_aggregate tests
Reviewers: teon.banek, buda
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1390
Summary:
Since we are moving from boost to Capnp for serialization, it makes
sense to keep all of the LogicalOperator classes in LCP format. This
will make it easier to generate Capnp code.
Depends on D1361
Reviewers: buda, mferencevic, msantl, dgleich, ipaljak, mculinovic, mtomic
Reviewed By: mtomic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1362
Summary:
Pointers in AstTreeStorage can point to same nodes. These nodes
should be serialized(saved) only once when serializing Ast tree.
Same goes for deserializing(loading).
When deserializing nodes, one should not use storage Create method,
because it will add node which isn't completely loaded to storage vector.
Therefore, one should use new Deserialize method which doesn't
add node to storage vector. Inserting node into storage vector
is defered until all node data is loaded.
Pass pointer to saved uids set, instead of storage
Delete unused set
Reviewers: teon.banek, msantl
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1344
Summary:
Added a new markdown file for concepts and a new test to test the edge
case with upper bound.
Reviewers: teon.banek, mtomic, dgleich, buda, ipaljak
Reviewed By: teon.banek, dgleich
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1366
Summary:
When commiting/aborting a transaction in tx master engine, make a two
phase commit to all workers so they can stop all futures and clear
transactional cache.
Reviewers: dgleich, florijan
Reviewed By: dgleich
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1320
Summary:
Before we used `utils::Future` only where it's created by our `ThreadPool`.
I suggest in this diff that we use it everywhere, it's a bit more defensive and
should not have any downsides.
Reviewers: msantl, teon.banek
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1316
Summary:
Add different priority VLOGs for distributed memgraph.
For level 3 you'll get logs for dispatching/consuming plans.
For level 4 you'll get logs for tx start/commit/abort, remote produce, remote
pull, remote result consume,
For level 5 there will be a log for each request/response made by the RPC
client.
Master log snippet P9
Worker log snippet P10
Reviewers: florijan, teon.banek
Reviewed By: florijan
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1296
Summary:
Remove "produce_" and "Produce" as prefix from all distributed stuff.
It's not removed in src/query/ stuff (operators).
Reviewers: dgleich
Reviewed By: dgleich
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1315
Summary:
Ensures that query plans are invalidated on the remote only when it's
guaranteed they will never be used on the master. This is done by
invalidating remote caches in the `CachedPlan` destructor.
There are two unplesant side-effects. First, an RPC call is made in an
object destructor. This is somewhat ugly, but not that different then
making an RPC call that must succeed in any other function. Note that
this does NOT slow down any query execution because the relevant
destructor is called by the skiplist garbage collector. The second ugly
side-effect is that in the unit test now we need to sleep to ensure the
skiplist GC destructs a cached plan before checking that it's
invalidated on the remote worker.
We might want to redesign this at some point.
Reviewers: teon.banek
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1302
Summary:
- Remove caches on workers as a result of plan expiration or race
during insertion.
- Extract caching functionality into a class.
- Minor refactor of Interpreter::operator()
- New RPC and test for it.
- Rename ConsumePlanRes to DispatchPlanRes for consistency, remove
return value as it's always true and never used.
- Interpreter is now constructed with a `GraphDb` reference. At the
moment only for reaching the `distributed::PlanDispatcher`, but in
the future we should probably use that primarily for planning.
I added a function to `PlanConsumer` that is only used for testing.
I prefer not doing this, but I felt this needed testing. I can remove
it now if you like.
Reviewers: teon.banek, msantl
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1292
Summary:
This fixes a bug where the planning would raise NotYetImplemented error,
due to preventing plan splitting while the operator is already on
master. For example, `MATCH (a), (b) CREATE (a)-[e:r]->(b) RETURN e`
would split the plan after Cartesian and before Create. Thus, the rest
of the plan would be on master, including the Accumulate before Produce.
We still can (and must) replace Accumulate with Synchronize but this
would fail due to unneeded check.
Reviewers: florijan, msantl
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1272
Summary:
I have not thoroughly thought this through, especially the worker
destruction (is it legit to abort all running tx?), but it's tested to
abort during remote pull, what we need.
Also I improved error handling for vertex deletion failure during
remote pull (@dgleich).
Reviewers: teon.banek, msantl, dgleich
Reviewed By: dgleich
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1263
Summary:
If the `db` says we need to abort we should abort. There are two
places (in both `PullRemote` and `PullRemoteOrderBy`) where the check is made.
The first one is on the very beginning of the `Pull` method and the second one
is in the loop that checks/waits for remote results.
Reviewers: teon.banek, florijan
Reviewed By: florijan
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1265
Summary:
Usually a single RPC call takes 50us, which is a lot lower than our
millisecond sleep. Calling sleep with less than a ms duration may not
be really supported (due to timer granularity), but this change should
cause some sub-ms sleep.
Reviewers: florijan, msantl, mferencevic
Reviewed By: florijan
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1258
Summary:
During distributed execution, OrderBy is split across workers and the
master gets to merge those results via PullRemoteOrderBy. Since this
operator may be an input to almost any other operator, virtual accessors
to `input` have been added in LogicalOperator.
Depends on D1221
Reviewers: florijan, msantl, buda
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1232
Summary:
Extracted `TypedValueVectorCompare` and `RemotePuller` from operators
so it can be reused.
The new `PullRemoteOrerBy` operator pulls one result from each worker and one
from master, relies on the fact that workers/master returned sorted results,
returns the next one in order, and pulls the source of that result to get the
next one.
Depends on D1215 that (at the moment) is still in review.
Reviewers: florijan, teon.banek
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1221
Summary:
- The expansion vertex gets created on the origin's worker
- The edge automatically gets created wherever necessary
- Vertex creation logic reuse
- End to end test
Reviewers: teon.banek, msantl
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1227
Summary:
- Add the said test
- Add `workerid` function
- Add random wait (enabled with flag) to `rpc::Client::Call` for network
latency simulation
Reviewers: buda, mferencevic, teon.banek
Reviewed By: buda, teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1222
Summary:
This may improve the query execution workload, because the Expand will
yield local results while it waits for remote ones. Note that we rely on
the fact that walking the graph produces results without any predetermined
order. Therefore, we can yield paths as we see fit. Enforcing the order
is done through OrderBy operator, and this change shouldn't affect that.
Reviewers: florijan, msantl, buda
Reviewed By: florijan
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1215
Summary:
- Add `database::GraphDb::GetWorkerIds()`
- Change `CreateNode` constructor API
- Make `CreateNode` distribute nodes uniformly over workers
Did not yet modify `CreateExpand`, coming in the next diff.
Reviewers: teon.banek
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1216
Summary:
This is still very much in progress. No advanced checks are done to
prevent planning unimplemented things. Basic Cartesian product should
work, for example `MATCH (a), (b) CREATE (a)-[:r]->(c)-[:r]->(b)`. But
anything more advanced may lead to undefined behaviour of the planner
and therefore execution. Use at your own risk!
Add ModifiedSymbols method to LogicalOperator
For planning Cartesian, we need information on which symbols are filled
by operator sub-trees. Currently, this is used to set symbols which
should be transferred over network. Later, they should be used to detect
whether filter expressions use symbols modified from Cartesian branches.
Then we will be able to ensure correct dependency of filters and their
behaviour.
Prepare DistributedPlan for multiple worker plans
Since Cartesian branches need to be split and handled by each worker, we
now dispatch multiple plans to workers.
Reviewers: florijan, msantl, buda
Reviewed By: florijan
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1208
Summary:
On the master cleanups are hooked directly into the transaction engine.
This is beneficial because the master might have bigger caches and we
want to clear them as soon as possible.
On the workers there is a periodic RPC call to the master about living
transactions, which takes care of releasing local caches. This is
suboptimal because long transactions will prevent cache GC (like with
data GC). It is however fairly simple.
Note that all cleanup is not done automatically and `RemotePull` has
been reduced accordingly. @msantl, please verify correctness and
consider if the code can be additionally simplified.
Reviewers: teon.banek, msantl
Reviewed By: msantl
Subscribers: pullbot, msantl
Differential Revision: https://phabricator.memgraph.io/D1202
Summary:
Pulls left op cursor and keeps the result, and then for each pull of
the right op cursor, adds all the left op results to produce a cartesian
product.
Reviewers: teon.banek, florijan
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1201
Summary:
Added a hidden flag, a int32 value, that represents the number of
milliseconds the thread that pulls remote results should sleep in case when the
local results are exhausted but there are still some remote results that are
pending.
Reviewers: teon.banek, florijan
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1198
Summary:
The same behaviour can now be achieved by passing a nullptr for input
operator.
Reviewers: florijan, msantl
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1195
Summary:
Distributed plan is now dispatched before inserting into a cache. This
prevents a race condition when another thread would take the cached plan
and assume it was dispatched, thus crashing during execution.
The fix introduces a possibility that the same plan may be distributed
and dispatched twice (with different plan IDs). This shouldn't raise an
issue during execution, because only one plan will be cached and thus
executed. Obviously, we still need to invalidate the unused plan from
worker's caches.
Reviewers: florijan, msantl
Reviewed By: florijan
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1178
Summary:
This is a basic take on planning distributed writes. Main logic is in handling
the Accumulate operator, which requires Synchronize operation if the results
are used after modification.
Tests for planning distributed writes have been added.
Reviewers: florijan, msantl
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1172
Summary:
This avoids the unnecessary work of storing symbols which can never be
read after an aggregation is complete. With regards to distributed, a
major benefit is gained in reducing what is transferred over the
network. Hopefully, this doesn't break some obscure case where we
actually needed to remember all used symbols.
Reviewers: florijan, mislav.bradac, msantl
Reviewed By: florijan
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1158
Summary:
Queries such as `RETURN 1` should only be run on a single machine. This
change assumes that a query should only be distributed if it contains at
least one `ScanAll` operator, i.e. a `MATCH` clause.
Reviewers: florijan, msantl
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1154
Summary:
Remote pulls can now be async. Note that
`RemotePullRpcClients::RemotePull` takes references to data structures
which should not be temporary in the caller. Still, maybe safer to make
copies?
Changed `RpcWorkerClients` API to make that possible.
Reviewers: dgleich, msantl, teon.banek
Reviewed By: msantl
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1150
Summary: PullRemoteCursor will pull all clients in a RoundRobin fashion until all clients are exhausted and there are no more results to return.
Reviewers: teon.banek, florijan
Reviewed By: florijan
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1147
Summary:
NOTE: This diff is still in progress. Many TODOs, lacking documentation
etc. But the main logic is there (some could be different), and it tests
out OK.
Reviewers: teon.banek, msantl, buda
Reviewed By: teon.banek
Subscribers: mferencevic, pullbot
Differential Revision: https://phabricator.memgraph.io/D1138
Summary:
Remove AdvanceCommand operator declaration.
Add query/plan/distributed source files.
Add hacked cloning of LogicalOperator via serialization.
Add virtual Default methods to CompositeVisitor.
Use BOOST_CLASS_EXPORT_KEY in ast.hpp and operator.hpp.
This is needed in a single binary to correctly serialize polymorphic
classes. The previous implementation worked because the memgraph_lib was
linked with test binaries, but nobody was actually serializing things
inside the single memgraph binary itself.
Print PullRemote symbols in tests/manual/query_planner
Add names to implicitly created aggregation symbols
Reviewers: florijan, msantl
Reviewed By: florijan
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1137
Summary:
With the added support for serialization, we should be able to transfer
plans across distributed workers.
The planner tests has been extended to test serialization. Operators
should be mostly tested, but the expression they contain aren't
completely. The quick solution is to use typeid for rudimentary
expression equality testing. The more involved solution of comparing
the expression tree for equality is the correct choice. It should be
done in the near future.
Reviewers: florijan, msantl
Reviewed By: florijan
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1122
Summary:
Other than the plan operators and the frame, we will need to pass the
generated symbol table to distributed workers.
Reviewers: florijan, msantl
Reviewed By: florijan
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1123
Summary:
Previously, we would have a `DCHECK` which crashes the application. This
was evident when testing a queries, such as:
MATCH (n) DELETE n SET n.prop = 42
Since the argument to update clauses is evaluated during execution, it
makes it very difficult to prevent such errors during semantic analysis.
For example:
MATCH (n)--(m) WITH collect(n) as ns, m
DETACH DELETE ns[m.prop] SET head(ns).prop = 42
Test query updates on deleted graph elements
Reviewers: florijan, dgleich
Reviewed By: florijan
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1114
Summary:
GraphDb is refactored to become an API exposing different parts
necessary for the database to function. These different parts can have
different implementations in SingleNode or distributed Master/Server
GraphDb implementations.
Interally GraphDb is implemented using two class heirarchies. One
contains all the members and correct wiring for each situation. The
other takes care of initialization and shutdown. This architecture is
practical because it can guarantee that the initialization of the
object structure is complete, before initializing state.
Reviewers: buda, mislav.bradac, dgleich, teon.banek
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1093
Summary:
Serialization of vertices and edges for distributed. Based on Boost
serialization. Threrefore moved TypedValue serialization from AST to
utils.
Reviewers: buda, dgleich, teon.banek
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1088
Summary:
If there was no plan caching, the CachedPlan would not survive
`Interpreter::operator()`, as it was not owned by the
`Interpreter::Result`. If there was caching, it could hapen that the
cache got invalidated while that plan was being interpreted (by
another thread) without that interpretation retaining ownership.
Also simplified code around this.
Reviewers: mislav.bradac, teon.banek
Reviewed By: mislav.bradac
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1084
Summary:
Although the first solution used cereal, the final implementation uses
boost. Since the cereal is still used in the codebase, compilation has
been modified to support multithreaded cereal.
In addition to serializing Ast classes, the following also needed to be
serialized:
* GraphDbTypes
* Symbol
* TypedValue
TypedValue is treated specially, by inlining the serialization code in
the Ast class, concretely PrimitiveLiteral.
Another special case was the Function Ast class, which now stores a
function name which is resolved to a concrete std::function on
construction.
Tests have been added for serialized Ast in
tests/unit/cypher_main_visitor
Reviewers: mferencevic, mislav.bradac, florijan
Reviewed By: mislav.bradac
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1067
