tzulitai opened a new pull request #116:
URL: https://github.com/apache/flink-statefun/pull/116
This is an initial support for state TTL for remote functions.
Users define state TTL in their YAML modules like so:
```
functions:
- function:
states:
- name: xxxx
ttlDurationMillis: 60000 # optional key
```
The current implementation has some limitations due to how we are
multiplexing remote function's user state in a single `PersistedTable` (see
FLINK-17954):
- The actual TTL being set will be the longest duration across all
registered state
- The only supported expiration mode now is `AFTER_READ_AND_WRITE`. That can
be added as a `ttlExpireMode` key in the YAML spec in the future.
## Main change log
- 8557fba to d27b143 some pre-req refactoring of `JsonModule` and
`HttpFunctionSpec`. This refactors out version-specific JSON parsing logic to a
separate class (`JsonModuleSpecParser`), and any concerns such as default
values for optional keys, to `HttpFunctionSpec`.
- fcc25ec to 91e22ef Introduce a `StateSpec` model class to
`HttpFunctionSpec` that captures the configured state ttl duration for each
state. The `StateSpec` is then used in `RequestReplyFunction` to correctly set
the state TTL on the state handles.
- 39a8745 to 0f7eb9a Adds a spec parser for v2.0 format, and wires it in to
allow the version to be recognized.
- f48c64e Refactors `JsonModuleTest` so that it runs tests for both v1.0 and
v2.0 format
## Verifying
- New `JsonModuleTest` covers basic parsing of the new format
- End-to-end tests don't fail imply that remote function state access is not
broken due to this change
- Manually adapted the Python greeter example to check that state is expired.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
