I was thinking about our earlier thread of a single threaded applying and 
sending keys to task queue for multithread processing etc.

Each thread gets a group id but they write to the same state machine DB and to 
scale up more threads with more group ids get added. (As part of that group ids 
may move around in unexpected ways) Then more servers to scale out with their 
own group ids.

The ingress protion then maps client requests key ranges to group ids and 
forwards it to the right place?

Bigger picture:
What I am trying to figure out is can I build a datastore that can be "easily" 
dynamically resized as  the volume of data grows.



________________________________
From: Tsz Wo Sze <[email protected]>
Sent: Sunday, March 6, 2022 3:28:29 AM
To: [email protected] <[email protected]>
Subject: Re: Change group id

Hi Asad,

Currently, we do not support changing group id.  What is your use case?

Tsz-Wo


On Sun, Mar 6, 2022 at 1:11 PM Asad Awadia 
<[email protected]<mailto:[email protected]>> wrote:
Can the group id be changed for a state machine that has data/transactions 
applied?

Will it create a new blank state machine to replicate? Or it will be able to 
still work on existing data?

Reply via email to