[MDEV-4824] userstats - wrong user statistics Created: 2013-07-29  Updated: 2013-11-12  Resolved: 2013-11-12

Status: Closed
Project: MariaDB Server
Component/s: None
Affects Version/s: 5.5.32
Fix Version/s: 5.5.34

Type: Bug Priority: Minor
Reporter: Patryk Pomykalski Assignee: Sergei Golubchik
Resolution: Fixed Votes: 0
Labels: user_statistics


 Description   

At least bytes_received can be wrong. In function update_global_user_stats_with_user:

user_stats->bytes_received+= (thd->org_status_var.bytes_received-
 
                                thd->start_bytes_received);

thd->start_bytes_received is set in do_command(), but it is not called if thd_prepare_connection() fails.

cpu_time looks strange also...

select USER, CONNECTED_TIME, BUSY_TIME, CPU_TIME, BYTES_RECEIVED from information_schema.user_statistics where user = 'cacti';
 
+-------+----------------+--------------------+-------------------+---------------------+
 
| USER  | CONNECTED_TIME | BUSY_TIME          | CPU_TIME          | BYTES_RECEIVED      |
 
+-------+----------------+--------------------+-------------------+---------------------+
 
| cacti |             46 | 0.3571019999999954 | 692552.4628808994 | 9223372036854775807 |
 
+-------+----------------+--------------------+-------------------+---------------------+



 Comments   
Comment by Patryk Pomykalski [ 2013-07-30 ]

Also found valgrind warnings when running ./mtr main.connect -valgrind --mysqld=-userstat

==7422== Thread 5:
==7422== Conditional jump or move depends on uninitialised value(s)
==7422== at 0x81F30EA: THD::update_all_stats() (sql_class.cc:1314)
==7422== by 0x822E580: dispatch_command(enum_server_command, THD*, char*, unsigned int) (sql_parse.cc:1462)
==7422== by 0x822C467: do_command(THD*) (sql_parse.cc:794)
==7422== by 0x833033D: do_handle_one_connection(THD*) (sql_connect.cc:1266)
==7422== by 0x832FE84: handle_one_connection (sql_connect.cc:1181)
==7422== by 0x858B6BF: pfs_spawn_thread (pfs.cc:1015)
==7422== by 0x404296D: start_thread (pthread_create.c:300)
==7422== by 0x42B4A4D: clone (clone.S:130)
==7422== Conditional jump or move depends on uninitialised value(s)
==7422== at 0x81F311A: THD::update_all_stats() (sql_class.cc:1316)
==7422== by 0x822E580: dispatch_command(enum_server_command, THD*, char*, unsigned int) (sql_parse.cc:1462)
==7422== by 0x822C467: do_command(THD*) (sql_parse.cc:794)
==7422== by 0x833033D: do_handle_one_connection(THD*) (sql_connect.cc:1266)
==7422== by 0x832FE84: handle_one_connection (sql_connect.cc:1181)
==7422== by 0x858B6BF: pfs_spawn_thread (pfs.cc:1015)
==7422== by 0x404296D: start_thread (pthread_create.c:300)
==7422== by 0x42B4A4D: clone (clone.S:130)

Comment by Elena Stepanova [ 2013-08-08 ]

The valgrind warning is reproducible as described on the current 5.5 tree built with BUILD/compile-pentium-valgrind-max-no-ndb. If it turns out to be unrelated to the wrong values problem, I can try to create a test case for that one.

Generated at Thu Feb 08 06:59:29 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.