deploy failed when last join

[vagetablechicken]

Error

https://github.com/4paradigm/OpenMLDB/actions/runs/4341576688/jobs/7581344528#step:5:1253

Error: [2000] new ttl type kLatestTime doesn't match the old ttl type kAbsoluteTime in table action
about table action:

last join `action` as `action_reqId` on `flattenRequest`.`reqId` = `action_reqId`.`reqId`

Error: [2000] new ttl type kAbsAndLat doesn't match the old ttl type kLatestTime in table flattenRequest
about table flattenRequest:

flattenRequest_user_id_eventTime_0s_14d_200 as (partition by `user_id` order by `eventTime` rows_range between 14d preceding and 0s preceding MAXSIZE 200)

Desc

Deploy sql will failed when it has last join. We can reproduce it by oneflow demo case in 0.7.2 docker.

docker run --name=openmldb -it ...
cat <<EOF | docker exec --interactive openmldb bash
/work/init.sh
/work/openmldb/bin/openmldb --zk_cluster=127.0.0.1:2181 --zk_root_path=/openmldb --role=sql_client < /work/oneflow_demo/sql_scripts/create_tables.sql
/work/openmldb/bin/openmldb --zk_cluster=127.0.0.1:2181 --zk_root_path=/openmldb --role=sql_client < /work/oneflow_demo/sql_scripts/deploy.sql
EOF

TODO

1. make deploy succeed
2. get the error if deploy failed in compose test
3. oneflow doc: fix the oneflow pip version and more info about cudnn

[vagetablechicken]

works: 0.6.4 0.6.7 0.7.0 0.7.1
0.7.2 and main failed on it.
v0.7.1...v0.7.2
No change in sql_cluster_router HandleDeploy?

[vagetablechicken]

You can test by

cat <<EOF | docker exec --interactive openmldb bash
/work/init.sh
/work/openmldb/bin/openmldb --zk_cluster=127.0.0.1:2181 --zk_root_path=/openmldb --role=sql_client < /work/oneflow_demo/sql_scripts/create_tables.sql
/work/openmldb/bin/openmldb --zk_cluster=127.0.0.1:2181 --zk_root_path=/openmldb --role=sql_client < /work/oneflow_demo/sql_scripts/deploy.sql
EOF

[vagetablechicken]

The root cause is 2197046. After it, deploy will get new index map(last join => latest 1), and can't pass in add index phase(lat and abs conflict).