Details
-
Bug
-
Status: Stalled (View Workflow)
-
Minor
-
Resolution: Unresolved
-
10.2(EOL)
-
None
Description
|
10.2 f2fe65106f, debug build, default configuration |
MariaDB [test]> show processlist;
|
+----+-------------+-----------------+------+---------+------+--------------------------+------------------+----------+ |
| Id | User | Host | db | Command | Time | State | Info | Progress | |
+----+-------------+-----------------+------+---------+------+--------------------------+------------------+----------+ |
| 1 | system user | | NULL | Daemon | NULL | InnoDB background thread | NULL | 0.000 | |
| 2 | system user | | NULL | Daemon | NULL | InnoDB background thread | NULL | 0.000 | |
| 4 | system user | | NULL | Daemon | NULL | InnoDB background thread | NULL | 0.000 | |
| 3 | system user | | NULL | Daemon | NULL | InnoDB background thread | NULL | 0.000 | |
| 5 | system user | | NULL | Daemon | NULL | InnoDB background thread | NULL | 0.000 | |
| 10 | root | localhost:39009 | test | Query | 0 | init | show processlist | 0.000 |
|
+----+-------------+-----------------+------+---------+------+--------------------------+------------------+----------+ |
6 rows in set (0.00 sec) |
I don't think they should be there
Attachments
Issue Links
- duplicates
-
-
- Closed
-
- is caused by
-
MDEV-5800 indexes on virtual (not materialized) columns
-
- Closed
-
- relates to
-
-
- Closed
-
-
-
- Closed
-
-
-
- Open
-
-
-
- Open
-
Activity
serg, fine with me, too.
marko, does not matter how many threads there are. Even if there is one purge, and one "shutdown proxy", it is still bad enough as user experience.
There is a kind-of-a middle ground between "client connections" and "system threads" – replication-related threads: Binlog Dump on master and all those Slave_IO, Slave_SQL and Slave_worker threads on the slave. Unlike other system threads, they actually show states which make sense to the user and might be important to them; AFAIR they can also be meaningfully killed under some circumstances. I am not really sure how it can and should be treated, I think if we lose them for the sake of cosmetical improvements, users might be unhappy about it.
The complete list of system threads:
- replication slave io and sql
- event scheduler and workers
- delayed insert
- binlog background thread
- InnoDB shutdown handler, purge coordinator and workers
Which of those should be hidden from SHOW PROCESSLIST? Only the last item? It's kind of inconsistent.
One point of listing every THD in the process list is, that a THD can hold locks, so it can participate in a deadlock or can block, say, ALTER TABLE or FLUSH TABLES WITH READ LOCK, so there might be a need to kill it.
wlad, negative id would be a separate feature, a easy way to distinguish visually internal from user threads. It cannot be done as a bug fix and is not part of this issue.
Skipping internal threads in PROCESSLIST and KILL can be done based on something like thd->is_internal_thread() (or whatever), I'll do that.