ccql/README.md

6.7 KiB

ccql

downloads

Concurrent, multi server MySQL client

ccql is a simple executable utility which executes a given set of queries on a given set of MySQL hosts in parallel.

Quick example:

echo "my.srv1.com my.srv2.com my.srv3.com" | ccql -q "show master status; select @@server_id" -u myuser -p 123456

Usage

Usage of ccql:
  -C string
        Credentials file, expecting [client] scope, with 'user', 'password' fields. Overrides -u and -p
  -H string
    	Hosts file, hostname[:port] comma or space or newline delimited format. If not given, hosts read from stdin
  -Q string
    	Query/queries input file
  -d string
      Default schema to use (default "information_schema")
  -h string
    	Comma or space delimited list of hosts in hostname[:port] format. If not given, hosts read from stdin
  -m uint
    	Max concurrent connections (default 32)
  -p string
    	MySQL password
  -q string
    	Query/queries to execute
  -t float
    	Connect timeout seconds
  -u string
    	MySQL username (default OS user)

Hosts input

You may provide a list of hosts in the following ways:

  • via -h my.srv1.com:3307 my.srv2.com my.srv3.com
  • via -H /path/to/hosts.txt
  • via stdin, as in echo "my.srv1.com:3307 my.srv2.com my.srv3.com" | ccql ...

Hostnames can be separated by spaces, commas, newline characters or all the above. They may indicate a port. The default port, if unspecified, is 3306

Queries input

You may provide a query or a list of queries in the following ways:

  • single query, -q "select @@global.server_id"
  • multiple queries, semicolon delimited: -q "select @@global.server_id; set global slave_net_timeout:=10"
  • single or mutiple queries from text file: -Q /path/to/queries.sql

Queries are delimited by a semicolon (;). The last query may, but does not have to, be terminated by a semicolon. Quotes are respected, up to a reasonable level. It is valid to include a semicolon in a quoted text, as in select 'single;query'. However ccql does not employ a full blown parser, so please don't overdo it. For example, the following may not be parsed correctly: select '\';\''. You get it.

Credentials input

You may provide credentials in the following ways:

  • via -u myusername -p mypassword (default username is your OS user; default password is empty)
  • via credentials file: -C /path/to/.my.cnf. File must be in the following format:
    [client]
    user=myuser
    password=mypassword
    

Execution

Hosts are executed in parallel, with up to 128 concurrent executions (otherwise more hosts are accepted but wait in queue). For each host, the set of queries executes sequentially. Error on any query terminates execution of that host. Errors are isolated to hosts; an error while connecting or executing on host1 should not affect execution on host2.

Output

There is only output generated for queries that provide an output, typically SELECT queries. Queries such as SET GLOBAL... or FLUSH BINARY LOGS or CREATE DATABASE ... do not generate and output.

Output is written to stdout. It is tab delimited. There is one output line per row returning from either query. The first printed token is the fully qualified hostname:port of the instance whose query output is printed. Remember that execution happens concurrently on multiple hosts. Output rows are therefore ordered arbitrarily in between hosts, though deterministically for any specific host. Other tokens are whatever columns were returned by the queries.

More examples

Some examples dealing with replication follow. Combining shell scripting we can have some real fun. For brevity, we assume /tmp/hosts.txt contains a list of servers, as follows:

echo "localhost:22293, localhost:22294, localhost:22295, localhost:22296" > /tmp/hosts.txt

(note that hosts can be separated by spaces, commas, newlines or any combination)

We also assume credentials are stored in /etc/ccql.cnf:

[client]
user=msandbox
password=msandbox

Warmup: select some stuff

cat /tmp/hosts.txt | ccql -C /etc/ccql.cnf -q "select @@global.server_id, @@global.binlog_format, @@global.version"

A sample output is:

localhost:22296	103	STATEMENT	5.6.28
localhost:22294	101	STATEMENT	5.6.28-log
localhost:22293	1	STATEMENT	5.6.28-log
localhost:22295	102	STATEMENT	5.6.28-log

The output is tab delimited.

Show only servers that are configured as replicas:

cat /tmp/hosts.txt | ccql -C /etc/ccql.cnf -q "show slave status" | awk '{print $1}'

Apply slave_net_timeout only on replicas:

cat /tmp/hosts.txt | ccql -C /etc/ccql.cnf -q "show slave status;" | awk '{print $1}' | ccql -C /etc/ccql.cnf -q "set global slave_net_timeout := 10"

Getting tired of typing ccql -C /etc/ccql.cnf? Let's make a shortcut:

alias ccql="ccql -C /etc/ccql.cnf"

Which servers are acting as masters to someone?

cat /tmp/hosts.txt | ccql -q "show slave status;" | awk -F $'\t' '{print $3 ":" $5}'

Of those, which are also replicating? i.e. act as intermediate masters?

cat /tmp/hosts.txt | ccql -q "show slave status;" | awk -F $'\t' '{print $3 ":" $5}' | sort | uniq | ccql -q "show slave status" | awk '{print $1}'

Set sync_binlog=0 on all intermediate masters:

cat /tmp/hosts.txt | ccql -q "show slave status;" | awk -F $'\t' '{print $3 ":" $5}' | sort | uniq | ccql -q "show slave status" | awk '{print $1}' | ccql -q "set global sync_binlog=0"

LICENSE

See LICENSE. ccql imports and includes 3rd party libraries, which have their own license. These are found under vendor.

Binaries, downloads

Find precompiled binaries for linux (amd64) and Darwin (aka OS/X, amd64) under Releases

Build

ccql is built with Go 1.6/1.7, and supports 1.5 with the Go 1.5 vendor directories, which requires setting GO15VENDOREXPERIMENT=1. Please see the build file

What's in a name?

ccql is an abbreviation for Concurrent Client for MySQL or something. We had a few iterations with the name but had to replace one and we were all like yeah and whoa and fun times. Eventually we came by this name which upset tomkrouper being "too much on the left-side of the keyboard when typing" and that settled the matter. Tom uses alias a='ccql'.

Notes

  • Credits to Domas Mituzas for creating pmysql. This project mostly reimplements pmysql and delivers it in an easy to redistribute format.

  • Pronounce "see-sequel"

  • This project is open to contributions. Generally speaking it should be kept small and simple.