Command-Line Options and Environment Variables
This tool does not use any environment variables. Any command-line options are simply passed through to the MySQL client tools when querying the database server.
How it Works
This tool works by connecting to a MySQL database server and querying it for status and configuration information. It saves these bits of data into files in /tmp, and then formats them neatly with awk and other scripting languages.
The tool interacts minimally with the server upon which it runs. It assumes that you'll run it on the same server you're inspecting, and therefore it assumes that it will be able to find the my.cnf configuration file, for example. However, it should degrade gracefully if this is not the case. Note, however, that its output does not indicate which information comes from the MySQL database and which comes from the host operating system, so it is possible for confusing output to be generated if you run the tool on one server and direct it to connect to a MySQL database server running on another server.
Fuzzy-Rounding
Many of the outputs from this tool are deliberately rounded to show their magnitude but not the exact detail. This is called fuzzy-rounding. The idea is that it doesn't matter whether a server is running 918 queries per second or 921 queries per second; such a small variation is insignificant, and only makes the output hard to compare to other servers. Fuzzy-rounding rounds in larger increments as the input grows. It begins by rounding to the nearest 5, then the nearest 10, nearest 25, and then repeats by a factor of 10 larger (50, 100, 250), and so on, as the input grows.
Example Usage
Simply execute the tool, passing any connection options that you wish to use for the underlying MySQL client tools. If you run the tool interactively, then it will prompt you to decide whether to dump and analyze the server's schema structures; if run non-interactively, it will assume No and skip the schema analysis step.
The following is an example report. Each section is followed by a discussion of its contents and meaning.
# Aspersa MySQL Summary Report ###############################
System time | 2010-09-22 15:33:38 UTC (local TZ: PDT -0700)
The first section simply shows the time the report was generated, in UTC time. This comes from the operating system where the tool is executed, not from the MySQL server.
# Instances ##################################################
Port Data Directory Socket
===== ========================== ======
3306 /var/lib/mysql/data /tmp/mysql.sock
The Instances report is generated from the output of ps and shows one line per server instance. Sometimes it is not able to detect some of the parameters such as the port number. From this point on, the report will be about a single MySQL server instance; if multiple instances are running, you might need to run multiple reports.
# Report On Port 3306 ########################################
User | root@localhost
Time | 2010-09-22 08:33:38 (PDT)
Hostname | db1.localdomain.com
Version | 5.1.47-rel11.2-log Percona
Built On | unknown-linux-gnu x86_64
Started | 2010-09-08 16:57 (up 13+15:35:47)
Databases | 5
Datadir | /var/lib/mysql/data/
Processes | 80 connected, 1 running
Replication | Is a slave, has 0 slaves connected
This section is a quick summary of the MySQL instance: version, uptime, and other very basic parameters. The Time output is generated from the MySQL server, unlike the system date and time printed earlier, so you can see whether the database and operating system's time matches.
# Processlist ################################################
Command COUNT(*) Working SUM(Time) MAX(Time)
------------------------------ -------- ------- --------- ---------
Connect 2 2 1000000 1000000
Query 1 1 0 0
Sleep 80 0 3000 150
User COUNT(*) Working SUM(Time) MAX(Time)
------------------------------ -------- ------- --------- ---------
readonly 80 0 0 0
root 1 1 0 0
system user 2 2 1000000 1000000
Host COUNT(*) Working SUM(Time) MAX(Time)
------------------------------ -------- ------- --------- ---------
10.10.18.253 80 0 0 0
2 2 1000000 1000000
localhost 1 1 0 0
db COUNT(*) Working SUM(Time) MAX(Time)
------------------------------ -------- ------- --------- ---------
lportal 80 0 0 0
NULL 3 3 1000000 1000000
State COUNT(*) Working SUM(Time) MAX(Time)
------------------------------ -------- ------- --------- ---------
80 0 0 0
Has read all relay log; waitin 1 1 0 0
NULL 1 1 0 0
Waiting for master to send eve 1 1 1000000 1000000
This section is a summary of the output from SHOW PROCESSLIST. Each sub-section is aggregated by a different item, which is shown as the first column heading. When summarized by Command, every row in SHOW PROCESSLIST is included, but otherwise, rows whose Command is Sleep are excluded from the SUM and MAX columns, so they don't skew the numbers too much.
The columns are the number of rows included, the number that are not in Sleep status, the sum of the Time column, and the maximum Time column. The numbers are fuzzy-rounded.
# Status Counters (Wait 10 Seconds) ##########################
Variable Per day Per second 10 secs
Aborted_clients 20
Binlog_cache_disk_use 500
Binlog_cache_use 450000 5 10
Bytes_received 12500000000 150000 350000
Bytes_sent 70000000000 800000 2500000
Com_begin 450000 5 10
Table_locks_immediate 60000000 700 1500
.......................... (many lines omitted) .......................
Threads_created 800
Uptime 90000 1 1
This section shows selected counters from two snapshots of SHOW GLOBAL STATUS, gathered approximately 10 seconds apart and fuzzy-rounded. It includes only items that are incrementing counters; it does not include absolute numbers such as the Threads_running status variable, which represents a current value, rather than an accumulated number over time.
The first column is the variable name, and the second column is the counter from the first snapshot divided by 86400, so you can see the magnitude of the counter's change per day. 86400 fuzzy-rounds to 90000, so the Uptime counter should always be about 90000.
The third column is the value from the first snapshot, divided by Uptime and then fuzzy-rounded, so it represents approximately how quickly the counter is growing per-second over the uptime of the server.
The third column is the incremental difference from the first and second snapshot, divided by the difference in uptime and then fuzzy-rounded. Therefore, it shows how quickly the counter is growing per second at the time the report was generated.
# Table cache ################################################
Size | 2048
Usage | 25%
This section shows the size of the table cache, followed by the percentage of the table cache in use. The usage is fuzzy-rounded.
# Query cache ################################################
query_cache_type | ON
Size | 0.0k
Usage | 0%
This section shows whether the query cache is enabled and its size, followed by the percentage of the cache in use. The usage is fuzzy-rounded.
# Schema #####################################################
Would you like to mysqldump -d the schema and analyze it? y/n y
There are 4 databases. Would you like to dump all, or just one?
Type the name of the database, or press Enter to dump all of them.
Database Tables Views SPs Trigs Funcs FKs Partn
lportal 249
maatkit 2
mysql 21
test 1
Database InnoDB MyISAM
lportal 249
maatkit 2
mysql 21
test 1
Database BTREE
lportal 936
maatkit 2
mysql 29
test
b v d i t l d b s t t m c d f s e l
i a a n i o o l m i e e h a l e n o
g r t t n n u o a m x d a t o t u n
i c e y g b b l e t i r e a m g
n h t i t l l s u t b
t a i n e e i t m l
r m t x n a i o
e t t m n b
p t
Database === === === === === === === === === === === === === === === === === ===
lportal 686 610 321 214 152 103 34 6 1 14 1 1 15 9
maatkit 2 1 5 4 25
mysql 8 1 4 19 2 4 3 7 2 58 6 77 5
test 1
If you select to dump the schema and analyze it, the tool will print the above section. This summarizes the number and type of objects in the database. It is generated by running mysqldump --no-data, not by querying the INFORMATION_SCHEMA, which can freeze a busy server.
You can choose not to dump the schema, to dump all of the databases, or to dump only a single named one, by specifying the appropriate options. In the example above, we are dumping all databases. If only a single one is dumped, its name will appear as {chosen}.
The first sub-report in the section is the count of objects by type in each database: tables, views, and so on. The second one shows how many tables use various storage engines in each database. The third sub-report shows the number of each type of indexes in each database.
The last section shows the number of columns of various data types in each database. For compact display, the column headers are formatted vertically, so you need to read downwards from the top. In this example, the first column is 'bigint' and the second column is 'varchar'.
All of the numbers in this portion of the output are exact, not fuzzy-rounded.
# Noteworthy Technologies ####################################
Full Text Indexing | No
Geospatial Types | No
Foreign Keys | No
Partitioning | No
SSL | No
Explicit LOCK TABLES | No
Delayed Insert | No
XA Transactions | No
NDB Cluster | No
Prepared Statements | Yes
This section shows some specific technologies used on this server. Some of them are detected from the schema dump performed for the previous sections; others can be detected by looking at SHOW GLOBAL STATUS.
# InnoDB #####################################################
Version | 1.0.8-11.2
Buffer Pool Size | 24.0G
Buffer Pool Fill | 100%
Buffer Pool Dirty | 0%
File Per Table | ON
Page Size | 16k
Log File Size | 2 * 256M = 512M
Log Buffer Size | 8M
Flush Method | O_DIRECT
Flush Log At Commit | 1
XA Support | ON
Checksums | ON
Doublewrite | ON
R/W I/O Threads | 4 4
I/O Capacity | 200
Thread Concurrency | 0
Concurrency Tickets | 500
Commit Concurrency | 0
Txn Isolation Level | REPEATABLE-READ
Oldest Transaction Seconds
Transaction States
75 not started
Tables Locked
Semaphore Waits
Semaphore Holders
This section shows important configuration variables for the InnoDB storage engine. The buffer pool fill percent and dirty percent are fuzzy-rounded. The last few lines are derived from the output of SHOW INNODB STATUS. It is likely that this output will change in the future to become more useful.
# MyISAM #####################################################
Key Cache | 384.0M
Pct Used | 20%
Unflushed | 0%
This section shows the size of the MyISAM key cache, followed by the percentage of the cache in use and percentage unflushed (fuzzy-rounded).
# Security ###################################################
Users | 12 users, 0 anon, 4 w/o pw, 4 old pw
Old Passwords | OFF
This section is generated from queries to tables in the mysql system database. It shows how many users exist, and various potential security risks such as old-style passwords and users without passwords.
# Binary Logging #############################################
Binlogs | 5
Zero-Sized | 0
Total Size | 4.9G
binlog_format | STATEMENT
expire_logs_days | 10
sync_binlog | 0
server_id | 2
binlog_do_db |
binlog_ignore_db |
This section shows configuration and status of the binary logs. If there are zero-sized binary logs, then it is possible that the binlog index is out of sync with the binary logs that actually exist on disk.
# Noteworthy Variables #######################################
Auto-Inc Incr/Offset | 1/1
default_storage_engine | 0
flush_time | 0
init_connect | 0
init_file | 0
sql_mode | 0
join_buffer_size | 128k
sort_buffer_size | 2M
read_buffer_size | 2M
read_rnd_buffer_size | 8M
bulk_insert_buffer | 0k
max_heap_table_size | 16M
tmp_table_size | 16M
max_allowed_packet | 1M
thread_stack | 256k
log | OFF
log_error | /var/lib/mysql/data/db1.localdomain.com.err
log_warnings | 1
log_slow_queries | ON
log_queries_not_using_indexes | OFF
log_slave_updates | ON
This section shows several noteworthy server configuration variables that might be important to know about when working with this server.
# Configuration File #########################################
Config File | Cannot autodetect, trying common locations
Config File | /etc/my.cnf
[client]
port = 3306
socket = /tmp/mysql.sock
[mysqld]
port = 3306
socket = /tmp/mysql.sock
language = /usr/share/mysql/english
skip-locking
key_buffer = 384M
..................... (many lines omitted) ......................
This section shows a pretty-printed version of the my.cnf file, with comments removed and with whitespace added to align things for easy reading. The tool tries to detect the my.cnf file by looking at the output of ps, and if it does not find the location of the file there, it tries common locations until it finds a file. Note: this file might not actually correspond with the server from which the report was generated. This can happen when the tool isn't run on the same server it's reporting on, or when detecting the location of the configuration file fails.