[
https://issues.apache.org/jira/browse/HDDS-14243?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ivan Andika updated HDDS-14243:
-------------------------------
Description:
HddsDispatcher#dispatchRequest is a single long complicated function which was
called by ContainerStateMachine#dispatchCommand which are called in multiple
places
* ContainerStateMachine#writeStateMachineData: For Write Pipeline V1 when
handling writeChunk
* ContainerStateMachine#readMachineData: For Write Pipeline V1 when reading
the cached writeChunk data before replicating to the followers
* ContainerStateMachine#getStreamDataChannel : For Streaming Write Pipeline
during the initial ContainerStateMachine#stream during stream init
* ContainerStateMachine#query: For read requests (this should be unused in
Ozone since Ozone uses normal gRPC read)
Even if HddsDispatcher#dispatchRequest call KeyValueHandler#handle which will
have a defer to handler request, the complicated HddsDispatcher#dispatchRequest
and lack of documentation makes it hard to know which handler is going to be
called.
For example, I'm trying to understand which
ContainerStateMachine#applyTransaction code path is going to be triggered when
client calls sendForward, but it's hard to determine since 1)
HddsDispatcher#dispatchRequest is complicated 2)
ContainerStateMachine#applyTransaction itself has more special logic (whether
to set BCSID, etc).
We need to simplify and document this.
was:
HddsDispatcher#dispatchRequest is a single long complicated function which was
called by ContainerStateMachine#dispatchCommand which are called in multiple
places
* ContainerStateMachine#writeStateMachineData: For Write Pipeline V1 when
handling writeChunk
* ContainerStateMachine#readMachineData: For Write Pipeline V1 when reading
the cached writeChunk data before replicating to the followers
* ContainerStateMachine#getStreamDataChannel : For Streaming Write Pipeline:
* ContainerStateMachine#query: For read requests (this should be unused)
Even if HddsDispatcher#dispatchRequest call KeyValueHandler#handle which will
have a defer to handler request, the complicated HddsDispatcher#dispatchRequest
and lack of documentation makes it hard to know which handler is going to be
called.
For example, I'm trying to understand which
ContainerStateMachine#applyTransaction code path is going to be triggered when
client calls sendForward, but it's hard to determine since 1)
HddsDispatcher#dispatchRequest is complicated 2)
ContainerStateMachine#applyTransaction itself has more special logic (whether
to set BCSID, etc).
We need to simplify and document this.
> Simplify and document HddsDispatcher#dispatch
> ---------------------------------------------
>
> Key: HDDS-14243
> URL: https://issues.apache.org/jira/browse/HDDS-14243
> Project: Apache Ozone
> Issue Type: Improvement
> Reporter: Ivan Andika
> Assignee: Ivan Andika
> Priority: Major
>
> HddsDispatcher#dispatchRequest is a single long complicated function which
> was called by ContainerStateMachine#dispatchCommand which are called in
> multiple places
> * ContainerStateMachine#writeStateMachineData: For Write Pipeline V1 when
> handling writeChunk
> * ContainerStateMachine#readMachineData: For Write Pipeline V1 when reading
> the cached writeChunk data before replicating to the followers
> * ContainerStateMachine#getStreamDataChannel : For Streaming Write Pipeline
> during the initial ContainerStateMachine#stream during stream init
> * ContainerStateMachine#query: For read requests (this should be unused in
> Ozone since Ozone uses normal gRPC read)
> Even if HddsDispatcher#dispatchRequest call KeyValueHandler#handle which will
> have a defer to handler request, the complicated
> HddsDispatcher#dispatchRequest and lack of documentation makes it hard to
> know which handler is going to be called.
> For example, I'm trying to understand which
> ContainerStateMachine#applyTransaction code path is going to be triggered
> when client calls sendForward, but it's hard to determine since 1)
> HddsDispatcher#dispatchRequest is complicated 2)
> ContainerStateMachine#applyTransaction itself has more special logic (whether
> to set BCSID, etc).
> We need to simplify and document this.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
