Summary:
Refactor in two ways. First, expose members without getters as we will
need most of them in distributed. And this was always the sensible thing
to do. Second, add storage type values to deltas. This is also a
sensible thing to do, and it will be very beneficial in distributed. We
didn't do it before because name<->value type mappings aren't guaranteed
to be the same after recovery. A task has been added to address this
(preserve mappings in durability).
Reviewers: dgleich, buda
Reviewed By: dgleich
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1167
Summary:
A hack worthy of young master Gleich. I *think* it's correct though, and
the tests pass. End-to-end cluster recovery testing will be written and
tried out by @mculinovic
Reviewers: dgleich, mculinovic
Reviewed By: dgleich
Subscribers: pullbot, mculinovic
Differential Revision: https://phabricator.memgraph.io/D1163
Summary:
Change `GraphDb` so it exposes index clients in the same
convention as other members.
Reviewers: dgleich, mculinovic
Reviewed By: mculinovic
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1159
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:
It is possible that we have a global address to resolve, for a graph
element that's local. Consider W1 expanding, getting data from W2,
expanding from there and getting data that is on W1. We then don't want
to do RPC from W1 to W1, but do a lookup directly.
Reviewers: dgleich
Reviewed By: dgleich
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1145
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:
Start removal of old logic
Remove more obsolete classes
Move Message class to RPC
Remove client logic from system
Remove messaging namespace
Move protocol from messaging to rpc
Move System from messaging to rpc
Remove unnecessary namespace
Remove System from RPC Client
Split Client and Server into separate files
Start implementing new client logic
First semi-working state
Changed network protocol layout
Rewrite client
Fix client receive bug
Cleanup code of debug lines
Migrate to accessors
Migrate back to binary boost archives
Remove debug logging from server
Disable timeout test
Reduce message_id from uint64_t to uint32_t
Add multiple workers to server
Fix compiler warnings
Apply clang-format
Reviewers: teon.banek, florijan, dgleich, buda, mtomic
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1129
Summary:
- End to end distributed GraphDb testing
- Refactors as necessary
- Basic RemoteCache for storing remote data
- RemoteDataRpc
As we are on a tight schedule, please let's focus on the essentials:
functionality and proper testing.
Reviewers: dgleich, teon.banek, buda
Reviewed By: dgleich
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1121
Summary:
A slight insanity here... I realized I will need to create
`GraphDbAccessor` instance (which need `&GraphDb`) within some members
of `::impl` classes. Within those classes I can pass `this` to those
members, if `this` is a valid `GraphDb`. Semantically it really is (at
the moment), but heirarchically it wasn't. This diff changes that.
`GraphDb` is now only an interface. `PublicBase` is the base for all
the public classes, `PrivateBase` for the `::impl` classes. Seems to
work.
Oh yes, another thing to keep in mind when doing this is that I should avoid
calling virtual functions in public classes (the motivation for the double
heirarchy). Before this diff the getters weren't virtual, now they are, so
I should have made all the appropriate changes in code as well.
Buda, was this a task I could have delegated to you or Cula?
Reviewers: teon.banek, dgleich, buda
Reviewed By: teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1120
Summary:
Adds worker id to snapshot and wal filename.
Adds a new worker_id flag to be used for recovering a worker with a distributed snapshot.
Adds worker_id field to snapshot to check for consistency.
Reviewers: florijan
Reviewed By: florijan
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1096
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:
No logic changes, just split `tx::MasterEngine` into
`tx::SingleNodeEngine` and `tx::MasterEngine`. This gives better
responsibility separation and is more appropriate now there is no
Start/Shutdown.
Reviewers: dgleich, teon.banek, buda
Reviewed By: dgleich, teon.banek
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1099
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:
This is a proposal on how the WAL recovery process can be implemented so
that Deltas aren't accumulated, but instead applied in the same order
they are written to the WAL.
I *believe* that the only additional requirement on the system are
atomic transaction Begin/Commit/Abort. By atomic I mean that they are
present in the WAL in exactly the same ordering like in the transaciton
engine, to ensure the same commitability of original and recovery
transactions.
This could be a requirement for HA recovery. It is desirable that WAL
and HA log become the same thing, and the recovery process too.
Reviewers: mtomic, dgleich, mislav.bradac
Reviewed By: mislav.bradac
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1068
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
Summary:
The distributed ID mapper is not yet utilised in GraphDb as those
changes are in D1060. Depending on landing order it will be added.
Reviewers: dgleich, mislav.bradac
Reviewed By: dgleich
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1064
Summary:
What's done:
- `RecordAccessor` can represent remote data
- `GraphDbAccessor` manages remote data
- Cleanup: different `EdgeAccessor lazyness (@dgleich: take a look), unused methods, documentation...
- `TODO` placeholders for remote implementation
What's not done:
- RPC and data transfer
- how exactly remote errors are handled
- not sure if any MVCC Record info for remote data should be tracked
- WAL and RPC Deltas properly handled (Gleich working on extracting `Wal::Op`)
This implementation should not break single-node execution, and should provide good abstractions and placeholders for distributed. Once that's satisfied, it should land.
Reviewers: dgleich, buda, mislav.bradac
Reviewed By: dgleich
Subscribers: dgleich, pullbot
Differential Revision: https://phabricator.memgraph.io/D1030
Summary: Operations are moved and renamed from WAL to a separate file in preparation for HA and distributed storage.
Reviewers: florijan, mtomic, mislav.bradac
Reviewed By: florijan
Subscribers: mislav.bradac, pullbot
Differential Revision: https://phabricator.memgraph.io/D1034
Summary:
The current idea is that the same MG binary can be used for single-node,
distributed master and distributed worker. The transactional engine in
the single-node and distributed master is the same: it determines the
transactional time and exposes all the "global" functionalities. In the
distributed worker the "global" functions must contact the master.
Reviewers: dgleich, mislav.bradac, buda
Reviewed By: dgleich
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1013
Summary: Because it will never be used, we already have replacements for it.
Reviewers: buda
Reviewed By: buda
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1016
Summary: Code simplification made possible by making `locks_` `mutable` in `tx::Transaction`.
Reviewers: dgleich, buda
Reviewed By: buda
Differential Revision: https://phabricator.memgraph.io/D1015
Summary:
This diff contains step 1:
- Remove clog exposure from tx::engine
- Reduce and cleanup tx::Engine API
All current functionality is kept, but the API is reduced. This is very
desirable because every function in tx::Engine will need to be
considered and implemented in both Master and Worker situations. The
less we have, the better.
Next step is exactly that: seeing how each of these functions behaves in
a distributed system and implementing accordingly.
Reviewers: dgleich, mislav.bradac, buda
Reviewed By: mislav.bradac
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D1008
Summary:
In preparation for distributed storage we need to have labels/properties/edgetypes uniquely identifiable by their ids, which will be global in near future.
The old design has to be abandoned because it's not possible to keep track of global labels/properties/edgetypes while they are local pointers.
Reviewers: mislav.bradac, florijan
Reviewed By: florijan
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D993
Summary: Vertex and Edge now use Address for storing connections to other Edges and Vertices, to support distributed storage.
Reviewers: mislav.bradac, dgleich, buda
Reviewed By: mislav.bradac, dgleich
Subscribers: pullbot
Differential Revision: https://phabricator.memgraph.io/D977
Summary:
My dear fellow Memgraphians. It's friday afternoon, and I am as ready to pop as WAL is to get reviewed...
What's done:
- Vertices and Edges have global IDs, stored in `VersionList`. Main storage is now a concurrent map ID->vlist_ptr.
- WriteAheadLog class added. It's based around buffering WAL::Op objects (elementraly DB changes) and periodically serializing and flusing them to disk.
- Snapshot recovery refactored, WAL recovery added. Snapshot format changed again to include necessary info.
- Durability testing completely reworked.
What's not done (and should be when we decide how):
- Old WAL file purging.
- Config refactor (naming and organization). Will do when we discuss what we want.
- Changelog and new feature documentation (both depending on the point above).
- Better error handling and recovery feedback. Currently it's all returning bools, which is not fine-grained enough (neither for errors nor partial successes, also EOF is reported as a failure at the moment).
- Moving the implementation of WAL stuff to .cpp where possible.
- Not sure if there are transactions being created outside of `GraphDbAccessor` and it's `BuildIndex`. Need to look into.
- True write-ahead logic (flag controlled): not committing a DB transaction if the WAL has not flushed it's data. We can discuss the gain/effort ratio for this feature.
Reviewers: buda, mislav.bradac, teon.banek, dgleich
Reviewed By: dgleich
Subscribers: mtomic, pullbot
Differential Revision: https://phabricator.memgraph.io/D958
