Hi everyone, we try to improve ALTER SCHEMA operation in shared-data cluster (3 FE, 3 CN, adls2). We found out, that the parameter alter_scheduler_interval_millisecond makes the difference - by default it's 10000 milliseconds, we changed to 100 milliseconds. But still it's not fast enough and takes around 0.300-0.400 seconds to add a new field to the empty table. And it's a problem if you try to add multiple fields one by one - you'll get

A schema change operation is in progress

error. I see two operations, that take too much time: • 0.100 seconds between these two logs in FE:

[SchemaChangeHandler.process():2043] finished to create schema change job: 1299569

[LakeTableAlterMetaJobBase.updateIndexTabletMeta():354] Sent update tablet metadata task. tableName=test_table partitionId=1299560 indexId=1299559 taskNum=1

• 0.125 seconds between these two logs in FE:

[LakeTableAlterMetaJobBase.updateNextVersion():396] LakeTableAlterMetaJob id: 1299569 update next version of partition: 1299560, commitVersion: 2

[LakeTableAlterMetaJobBase.updateVisibleVersion():408] partitionVisibleVersion=2 commitVersion=2

Anything we can do with it?