Make max_connections and innodb_buffer_pool_size configurable
through soft params instead of hard.

Change-Id: Id0bb8f6b6091289275864bf4dd6a6890d11b4782

Change-Id: Id77cd6da330fe986844b3331edd514819471e67a

We need to be sure that enabled param is defined and set to True
during a state applying.

Change-Id: I19cd4c834f254d76d6b521d983d05b227c048065

The buffer_pool_size parameter should be a integer value when followed
by a memory unit (eg '1024M', not '1024.1M'). Otherwise the MySQL
doesn't start properly.

Change-Id: I5dfad49a8e44a40ae87c5e8e0b033dcd73dc2837

Change-Id: I4de34ad6db6805235966ebcca6fe43ea1faaea42

Usage:

  galera:
    master:
      innodb_buffer_pool_size: 1024M
    slave:
      innodb_buffer_pool_size: 1024M

If innodb_buffer_pool_size is not set then 35% of the available
ram will be allocated for the buffer pool.

Change-Id: Iaffa3f5b4ded8bc838f9d707916e360ed873c13b

Change-Id: I2b6abf51793fa6179c3a6477a752243db6ac6e73

Change-Id: I9daada1cf87873d602d415642bbc4bd32d8e2d52

command called by service.dead state:
systemd-run --scope systemctl stop mysql.service

More common way is to shut down service using known credentials

Change-Id: I2abe7eeea850333139587347039b1de8a6ea486f

Optimize 'noservices' condition

Change-Id: Ia6873c44f4e5b179a6c4a545fbf4b45c978812ea

The tag is equal to the address of the server which is quite useless
when using a Unix socket address.

Change-Id: Ia87c30f8ff2176f923982d70b712b10ec6e97f56
Depends-On: I538dbb24602a8ad4aece0a74445db47810099bb9

Change-Id: I34d6d8146489bab30e5ee621e3b8d01fe7ff819b

Change-Id: Iec2d27cc4fdab9ba7772e9c26be66ba54f5aadfc
Depends-On: Ice4f38abfb0d870c56763dfeda2140032fa752a1

Fix case of systemd [Service] unit

Added Kitchen tests and travis

Change-Id: I6047d632a2bfeb6525ac60092436ccd5334671e4

Change-Id: I9638c66d1a204a54e25144c4dfe43d5cf5ba135c

PROD-10495 Fixing file match for correct targeting

Change-Id: I2856f50977b1959801d66e7bc73fe54d24d43801

It comes as dependency from main wsrep packages and it differs
between versions of OS.

Change-Id: I2b50edaa19e52009766bca9fa31c566f59ff9319

Because the current timeout interval is too short in some environments.
The change also adds information messages to bootstrap scripts

Change-Id: I295da792349f5864a484e3eb318342dcc31fffab

Change-Id: I91f55d599f5291f34bfa81f55dafd6734606e1cc

Galera master installation hangs on installation
mysql-server-wsrep-5.6. It seems that log_error is not working,
because of next message: "mysqld_safe Can't log to error log and
syslog at the same time. Remove all --log-error configuration
options for --syslog to take effect." Using "syslog = yes" instead
of log_error makes it works without any other changes.

Change-Id: I39a8a87ed048557c19f3843b97c75469385e4978
Closes-bug: PROD-9065

Provide grains for building the main dashboard

disable-sensu

Split the mysql alarm cluster

Fix mysql with adding ulimit -n 102400

Bug: After deployment of linux formulas and salt states
/etc/security/limits.conf are not applying, as result connections
to mysql failed with error (pymysql.err.OperationalError)
(1040, u'Too many connections')

Solution: create mysql.overide, where ulimit is set to 102400

Closed-bug: PROD-8427

Change cluster policy to availability_of_members

Because the cluster isn't managed by Pacemaker, using
majority_of_members is misleading.

Stacklight

Put Grafana dashboards into their own directory

Merge master into stacklight

Add alarms and alarm clusters