[Bug] drop table column fail

[Kree0]

Describe the bug

Version: cnosdb 2.3.4, revision 508f669
Mode: singleton | 3meta+2tskv

public ❯ create table t1(a bigint, b bigint);
Query took 0.006 seconds.
public ❯ ALTER TABLE t1 drop b;
422 Unprocessable Entity, details: {"error_code":"050014","error_message":"Unreachable"}

Stack backtrace:
   0: anyhow::error::<impl anyhow::Error>::msg
   1: anyhow::__private::format_err
   2: client::ctx::SessionContext::sql::{{closure}}
   3: client::exec::exec_and_print::{{closure}}
   4: client::exec::exec_from_repl::{{closure}}
   5: cnosdb_cli::main::{{closure}}
   6: cnosdb_cli::main
   7: std::sys_common::backtrace::__rust_begin_short_backtrace
   8: std::rt::lang_start::{{closure}}
   9: core::ops::function::impls::<impl core::ops::function::FnOnce<A> for &F>::call_once
             at ./rustc/39f2657d1101b50f9b71ae460b762d330cc8426b/library/core/src/ops/function.rs:287:13
  10: std::panicking::try::do_call
             at ./rustc/39f2657d1101b50f9b71ae460b762d330cc8426b/library/std/src/panicking.rs:487:40
  11: std::panicking::try
             at ./rustc/39f2657d1101b50f9b71ae460b762d330cc8426b/library/std/src/panicking.rs:451:19
  12: std::panic::catch_unwind
             at ./rustc/39f2657d1101b50f9b71ae460b762d330cc8426b/library/std/src/panic.rs:140:14
  13: std::rt::lang_start_internal::{{closure}}
             at ./rustc/39f2657d1101b50f9b71ae460b762d330cc8426b/library/std/src/rt.rs:148:48
  14: std::panicking::try::do_call
             at ./rustc/39f2657d1101b50f9b71ae460b762d330cc8426b/library/std/src/panicking.rs:487:40
  15: std::panicking::try
             at ./rustc/39f2657d1101b50f9b71ae460b762d330cc8426b/library/std/src/panicking.rs:451:19
  16: std::panic::catch_unwind
             at ./rustc/39f2657d1101b50f9b71ae460b762d330cc8426b/library/std/src/panic.rs:140:14
  17: std::rt::lang_start_internal
             at ./rustc/39f2657d1101b50f9b71ae460b762d330cc8426b/library/std/src/rt.rs:148:20
  18: main
  19: <unknown>
  20: __libc_start_main
  21: _start

To Reproduce

No response

Expected behavior

No response

Additional context

No response

[zipper-meng]

There is a bug that database in TsKv: Engine can be detected only after we wrote something to it:

cnosdb/tskv/src/kvcore.rs

Lines 959 to 986 in 6765f19

	async fn drop_table_column(
	&self,
	tenant: &str,
	database: &str,
	table: &str,
	column_name: &str,
	) -> Result<()> {
	// TODO(zipper): Store this action in WAL.
	let db = self.get_db(tenant, database).await?;
	let schema =
	db.read()
	.await
	.get_table_schema(table)?
	.ok_or_else(|| SchemaError::TableNotFound {
	database: database.to_string(),
	table: table.to_string(),
	})?;
	let column_id = schema
	.column(column_name)
	.ok_or_else(|| SchemaError::FieldNotFound {
	database: database.to_string(),
	table: table.to_string(),
	field: column_name.to_string(),
	})?
	.id;
	self.delete_columns(db, table, &[column_id]).await?;
	Ok(())
	}

Line 967 self.get_db() returned Err(DatabaseNotFound), then upstream printed the log and returned Unreachable to user.

[Kree0]

It's solved in cnosdb 2.4.1, revision cc1af6a

public ❯ create table t1(a bigint, b bigint);
Query took 0.027 seconds.
public ❯ ALTER TABLE t1 drop b;
Query took 0.029 seconds.