Hi, Rucha! On Jul 22, Rucha Deodhar wrote:

revision-id: ae1f4e4d727 (mariadb-10.5.11-1-gae1f4e4d727) parent(s): dc82effa5df author: Rucha Deodhar committer: Rucha Deodhar timestamp: 2021-06-21 20:24:43 +0530 message:

MDEV-23836: Assertion `! is_set() || m_can_overwrite_status' in Diagnostics_area::set_error_status (interrupted ALTER TABLE under LOCK)

Analysis: KILL_QUERY is not ignored when local memory used exceeds maximum session memory. Hence the query proceeds, OK is sent and we end up reopening tables that are marked for reopen. During this, kill status is eventually checked and assertion failure happens during trying to send error message because OK has already been sent. Fix: If local memory used exceeds maximum session memory and query is killed before it has completed execution, return TRUE.

diff --git a/sql/sql_table.cc b/sql/sql_table.cc index f58ce8f997d..0b722984fca 100644 --- a/sql/sql_table.cc +++ b/sql/sql_table.cc @@ -6489,6 +6489,12 @@ handle_if_exists_options(THD *thd, TABLE *table, Alter_info *alter_info, ER_BAD_FIELD_ERROR, ER_THD(thd, ER_BAD_FIELD_ERROR), sql_field->change.str, table->s->table_name.str); + /* + if thread is killed (for example because local memory used execeeds + maximum session memory used), return TRUE and do rollback (done later). + */ + if (thd->killed) + DBUG_RETURN(true); it.remove();

Looks like a completely arbitrary place to put a check in. What if the kill signal arrives later? Regards, Sergei VP of MariaDB Server Engineering and security@mariadb.org