[OAK-7932] A distributed node store for the cloud

• Type: Wish
• Status: Resolved
• Priority: Minor
• Resolution: Abandoned
• Component/s: segment-tar
• Assignee: Axel Hanikel
• Reporter: Axel Hanikel
• Created: 03/Dec/18 12:32
• Updated: 15/Dec/22 08:51
• Resolved: 15/Dec/22 08:51

Description

Outline

This issue documents some proof-of-concept work for adapting the segment tar nodestore to a
distributed environment. The main idea is to adopt an actor-like model, meaning:

*   Communication between actors (services) is done exclusively via messages.
*   An actor (which could also be a thread) processes one message at a time, avoiding sharing
    state with other actors as far as possible.
*   Nodestates are kept in RAM and are written to external storage lazily only for disaster recovery.

• A nodestate is identified by a uuid, which in turn is a hash on its serialised string representation.
  *   As RAM is a very limited resource, different actors own their share of the total uuid space.
  *   An actor might also cache a few nodestates which it does not own but which it uses often (such as
      the one containing the root node)

Implementation

The first idea was to use the segment node store, and ZeroMQ for communication because it seems to be a high-quality and
easy to use implementation. A major drawback is that the library is written in C and the Java
library which does the JNI stuff seems hard to set up and did not work for me. There is a native
Java implementation of the ZeroMQ protocol, aptly called jeromq, which seems to work well so far. This approach is probably not
pursued because due to the nature of how things are stored in segments, they are hard to cache (it seems like a large part of the repository
will eventually end up in the cache).

A second implementation, was a simple
nodestore implementation which was kind of a dual to the segment store in the sense that it is on the other end
of the compactness spectrum. The segment store is very dense and avoids duplication whereever possible.
The nodestore in this implementation, however, is quite redundant: Every nodestate gets its own UUID (a hash of the serialised
nodestate) and is saved together with its properties, similar to the document node store.

Here is what a serialised nodestate looks like:

begin ZeroMQNodeState
begin children
allow	856d1356-7054-3993-894b-a04426956a78
end children
begin properties
jcr:primaryType <NAME> = rep:ACL
:childOrder <NAMES> = [allow]
end properties
end ZeroMQNodeState

This verbose format is good for debugging but expensive to generate and parse, so it may be replaced with a binary format at some point. But it shows how child nodestates are referenced and how properties are represented. Binary properties (not shown here) are represented by a reference to the blob store.

The redundancy (compared with the segment store with its fine-grained record structure) wastes space, but on the other hand garbage
collection (yet unimplemented) is easier because there is no segment that needs to be rewritten to get rid of data that is no
longer referenced; unreferenced nodes can just be deleted. This implementation still has bugs, but being much simpler
than the segment store, it can eventually be used to experiment with different configurations and examine their
performance.

 A third implementation, at https://github.com/ahanikel/jackrabbit-oak/tree/zeromq-nodestore is based on the second one but uses a more compact serialisation format, which records the differences between a new nodestate and its base state, similar to what .compareAgainstBaseState() produces in code. It looks like this:

cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 n: 823f2252-db37-b0ca-3f7e-09cd073b530a 00000000-0000-0000-0000-000000000000
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 n+ root 00000000-0000-0000-0000-000000000000
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 n+ checkpoints 00000000-0000-0000-0000-000000000000
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 n+ blobs 00000000-0000-0000-0000-000000000000
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 n!
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 journal golden 823f2252-db37-b0ca-3f7e-09cd073b530a
cc97d19b-efa9-4f64-a5e9-a94d9418f271-151 checkpoints golden 00000000-0000-0000-0000-000000000000
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 n: 026cb997-d3f7-2c9f-a5ad-d7396b86c267 00000000-0000-0000-0000-000000000000
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 p+ :clusterId <STRING> = 0cb833ee-2678-4383-93dd-fe993cbfd56a
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 n!
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 n: 3ceb887c-0875-1171-94d2-df9bc8a57edd 00000000-0000-0000-0000-000000000000
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 n+ :clusterConfig 026cb997-d3f7-2c9f-a5ad-d7396b86c267
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 n!
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 n: ce8ace33-9d80-d6e1-c4ba-f0882f75e9c9 823f2252-db37-b0ca-3f7e-09cd073b530a
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 n^ root 3ceb887c-0875-1171-94d2-df9bc8a57edd 00000000-0000-0000-0000-000000000000
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 n!
cc97d19b-efa9-4f64-a5e9-a94d9418f271-26 journal golden ce8ace33-9d80-d6e1-c4ba-f0882f75e9c9
cc97d19b-efa9-4f64-a5e9-a94d9418f271-38 b64+ 516f4f64-4e5c-432e-b9bf-0267c1e96541
cc97d19b-efa9-4f64-a5e9-a94d9418f271-38 b64d ClVVRU5DT0RFKDEpICAgICAgICAgICAgICAgQlNEIEdlbmVyYWwgQ29tbWFuZHMgTWFudWFsICAg
cc97d19b-efa9-4f64-a5e9-a94d9418f271-38 b64d ICAgICAgICAgICBVVUVOQ09ERSgxKQoKTghOQQhBTQhNRQhFCiAgICAgdQh1dQh1ZAhkZQhlYwhj
cc97d19b-efa9-4f64-a5e9-a94d9418f271-38 b64d bwhvZAhkZQhlLCB1CHV1CHVlCGVuCG5jCGNvCG9kCGRlCGUgLS0gZW5jb2RlL2RlY29kZSBhIGJp
cc97d19b-efa9-4f64-a5e9-a94d9418f271-38 b64d bmFyeSBmaWxlCgpTCFNZCFlOCE5PCE9QCFBTCFNJCElTCFMKICAgICB1CHV1CHVlCGVuCG5jCGNv
cc97d19b-efa9-4f64-a5e9-a94d9418f271-38 b64d CG9kCGRlCGUgWy0ILW0IbV0gWy0ILW8IbyBfCG9fCHVfCHRfCHBfCHVfCHRfCF9fCGZfCGlfCGxf
cc97d19b-efa9-4f64-a5e9-a94d9418f271-38 b64d CGVdIFtfCGZfCGlfCGxfCGVdIF8Ibl8IYV8IbV8IZQogICAgIHUIdXUIdWQIZGUIZWMIY28Ib2QI
cc97d19b-efa9-4f64-a5e9-a94d9418f271-38 b64d ZGUIZSBbLQgtYwhjaQhpcAhwcwhzXSBbXwhmXwhpXwhsXwhlIF8ILl8ILl8ILl0KICAgICB1CHV1
cc97d19b-efa9-4f64-a5e9-a94d9418f271-38 b64d CHVkCGRlCGVjCGNvCG9kCGRlCGUgWy0ILWkIaV0gLQgtbwhvIF8Ib18IdV8IdF8IcF8IdV8IdF8I
cc97d19b-efa9-4f64-a5e9-a94d9418f271-38 b64d X18IZl8IaV8IbF8IZSBbXwhmXwhpXwhsXwhlXQo=
cc97d19b-efa9-4f64-a5e9-a94d9418f271-38 b64!

Every event starts with a universal thread identifier (a random UUID for the nodestore plus the threadId), followed by an operator: n: starts a nodestate. n+ and n^ both reference a nodestate which has been defined before. The first one says that a node has been created, the second one means an existing node has been changed, indicating the nodestate's uuid and the uuid of the basestate. This is slightly different from the format described in OAK-7849 which doesn't record the uuids. b64+ adds a new base64 encoded blob with the given uuid followed by the encoded data.