Percona Server with XtraDB

in slow query log administrator commands seem to inherit counters

Reported by Aurimas Mikalauskas on 2011-03-09
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Percona Server
Undecided
Unassigned

Bug Description

Apparently the application I'm working with is sending a ping command every so often which seems to be trashing slow query log with stats that later will break averages when aggregated. Here's few samples from slow query:

# User@Host: db[db] @ [192.168.100.33]
# Thread_id: 140286 Schema: db Last_errno: 0 Killed: 0
# Query_time: 0.007042 Lock_time: 0.000054 Rows_sent: 1 Rows_examined: 3 Rows_affected: 0 Rows_read: 0
# Bytes_sent: 688 Tmp_tables: 0 Tmp_disk_tables: 0 Tmp_table_sizes: 0
# InnoDB_trx_id: 9FB2839
# QC_Hit: No Full_scan: No Full_join: No Tmp_table: No Tmp_table_on_disk: No
# Filesort: Yes Filesort_on_disk: No Merge_passes: 0
# InnoDB_IO_r_ops: 1 InnoDB_IO_r_bytes: 16384 InnoDB_IO_r_wait: 0.006665
# InnoDB_rec_lock_wait: 0.000000 InnoDB_queue_wait: 0.000000
# InnoDB_pages_distinct: 8
SET timestamp=1299689295;
SELECT query goes here...
# User@Host: db[db] @ [192.168.100.33]
# Thread_id: 140286 Schema: db Last_errno: 0 Killed: 0
# Query_time: 0.000008 Lock_time: 0.000000 Rows_sent: 1 Rows_examined: 3 Rows_affected: 0 Rows_read: 0
# Bytes_sent: 699 Tmp_tables: 0 Tmp_disk_tables: 0 Tmp_table_sizes: 0
# InnoDB_trx_id: 9FB2839
# QC_Hit: No Full_scan: No Full_join: No Tmp_table: No Tmp_table_on_disk: No
# Filesort: Yes Filesort_on_disk: No Merge_passes: 0
# InnoDB_IO_r_ops: 1 InnoDB_IO_r_bytes: 16384 InnoDB_IO_r_wait: 0.006665
# InnoDB_rec_lock_wait: 0.000000 InnoDB_queue_wait: 0.000000
# InnoDB_pages_distinct: 8
SET timestamp=1299689295;
# administrator command: Ping;

# User@Host: db[db] @ [192.168.100.33]
# Thread_id: 138988 Schema: db Last_errno: 0 Killed: 0
# Query_time: 0.110581 Lock_time: 0.000057 Rows_sent: 1 Rows_examined: 2 Rows_affected: 0 Rows_read: 0
# Bytes_sent: 688 Tmp_tables: 0 Tmp_disk_tables: 0 Tmp_table_sizes: 0
# InnoDB_trx_id: 9FB05E4
# QC_Hit: No Full_scan: No Full_join: No Tmp_table: No Tmp_table_on_disk: No
# Filesort: Yes Filesort_on_disk: No Merge_passes: 0
# InnoDB_IO_r_ops: 2 InnoDB_IO_r_bytes: 32768 InnoDB_IO_r_wait: 0.109936
# InnoDB_rec_lock_wait: 0.000000 InnoDB_queue_wait: 0.000000
# InnoDB_pages_distinct: 9
use db;
SET timestamp=1299688570;
SELECT query
# User@Host: db[db] @ [192.168.100.33]
# Thread_id: 138988 Schema: db Last_errno: 0 Killed: 0
# Query_time: 0.000007 Lock_time: 0.000000 Rows_sent: 1 Rows_examined: 2 Rows_affected: 0 Rows_read: 0
# Bytes_sent: 699 Tmp_tables: 0 Tmp_disk_tables: 0 Tmp_table_sizes: 0
# InnoDB_trx_id: 9FB05E4
# QC_Hit: No Full_scan: No Full_join: No Tmp_table: No Tmp_table_on_disk: No
# Filesort: Yes Filesort_on_disk: No Merge_passes: 0
# InnoDB_IO_r_ops: 2 InnoDB_IO_r_bytes: 32768 InnoDB_IO_r_wait: 0.109936
# InnoDB_rec_lock_wait: 0.000000 InnoDB_queue_wait: 0.000000
# InnoDB_pages_distinct: 9
SET timestamp=1299688570;
# administrator command: Ping;

Server 5.1.55-rel12.5-log Percona Server (GPL), 12.5, Revision 200

Stewart Smith (stewart) wrote :

Are you saying that having Admin commands (such as ping) is the problem? Your aggregation should probably exclude these (or show them separately). Please let me know if I'm missing something.

Stewart Smith (stewart) wrote :

i.e. I don't think mere existence of Ping in query log is a bug, but rather a bug in producing good aggregation statistics.

I can be persuaded, but currently I don't consider this a bug.

Changed in percona-server:
status: New → Invalid

Steward,

sure I can just skip these statements, I wrote about it here:

http://www.mysqlperformanceblog.com/2011/03/28/maatkits-mk-query-digest-filters/

And I'm not saying having admin commands is the problem, the problem is that counters seem to be showing values of a previous statement, for example:

# InnoDB_IO_r_ops: 2 InnoDB_IO_r_bytes: 32768 InnoDB_IO_r_wait: 0.109936
# InnoDB_pages_distinct: 9
...
SELECT query
# InnoDB_IO_r_ops: 2 InnoDB_IO_r_bytes: 32768 InnoDB_IO_r_wait: 0.109936
# InnoDB_pages_distinct: 9
...
# administrator command: Ping;

Why would Ping command issue 2 read operations, read 32kbytes and look at 9 distinct pages? Why would it even have InnoDB stats at all?

To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers