The check currently prioritizes the credentials in the defaults extra files rather than those specified in the backupsets configuration files. If Holland is pointed to a remote instance (e.g. DBaaS) rather than the local instance and the defaults extra file points to the local mysql instance, then the check gives the status on the local instance rather than the remote that the backupsets configuration file points to.

Holland has the ability to have multiple backupset configurations active at once. The monitoring check only has the ability to check one. In addition the backupset is fixed at the "default" backupset when no backupset is given on the check line rather than reading the active backupsets out of the Holland configuration file.

Plugin will exit with a 1 status causing attached notifications to fire when client.log is empty since the script doesn't expect that.

Scout has such a plugin, written in Ruby, which might be helpful as a starting point.

We've had reports of false positives for chef_node_checking.py where no errors show up in the chef client logs, but alerts are triggered.

[root@hcluster1-db2 plugins]# ./holland_mysqldump.py
status success holland checked
metric log_age int64 138
metric dump_age int64 139
metric error_count int64 0
metric first_error string none
metric last_error string none
metric sql_creds_exist string false
metric sql_ping_succeeds string true
metric sql_status_succeeds string true
[root@hcluster1-db2 plugins]#

Although, we can see my credentials file is defined as ~/my.cnf.

[root@hcluster1-db2 plugins]# grep ^defaults /etc/holland/backupsets/default.conf
defaults-extra-file = ~/.my.cnf,
[root@hcluster1-db2 plugins]#

And this file does indeed exist.

[root@hcluster1-db2 plugins]# stat ~/.my.cnf
  File: ‘/root/.my.cnf’
  Size: 87          Blocks: 8          IO Block: 4096   regular file
Device: ca01h/51713d    Inode: 786463      Links: 1
Access: (0644/-rw-r--r--)  Uid: (    0/    root)   Gid: (    0/    root)
Access: 2016-05-18 18:06:54.239755077 -0500
Modify: 2016-05-18 18:06:54.239755077 -0500
Change: 2016-05-18 18:06:54.239755077 -0500
 Birth: -
[root@hcluster1-db2 plugins]#

Switching this to /root/.my.cnf file seems to be a good work around to make this detect the file properly.

[root@hcluster1-db2 plugins]# grep ^defaults /etc/holland/backupsets/default.conf
defaults-extra-file = /root/.my.cnf,
[root@hcluster1-db2 plugins]# ./holland_mysqldump.py
status success holland checked
metric log_age int64 335
metric dump_age int64 336
metric error_count int64 0
metric first_error string none
metric last_error string none
metric sql_creds_exist string true
metric sql_ping_succeeds string true
metric sql_status_succeeds string true
[root@hcluster1-db2 plugins]#

Although, as a more permanent solution, I ended up modifying line 144 from:

                if os.access(f, os.F_OK):

To:

                if os.access(os.path.expanduser(f), os.F_OK):

Pull request to follow.

By default, /root/.my.cnf is used as the credentials for the mysql ping that Holland monitoring performs. This means if a user changes the root password, monitoring fails, even though Holland may still run without issue.

In discussing how this might best be fixed, the two ideas where to simplify the ping check (and not pass the --defaults-file at all, just the Holland user/pass) or to remove the check and use https://github.com/racker/rackspace-monitoring-agent-plugins-contrib/blob/master/mysql_ping.py to monitor MySQL in real time instead of trying to do everything in the one plugin.

The ping and status checks don't have the option to specify the hosts being checked.

I installed the mongodb plugin. Is it supposed to show up in mycloud dashboard, or will it only be present on the API?

Also what is the proper path to install the plugin? I am using debian squeeze. Thanks.

Based on feedback from new user who was pointed to this place, we should update the document to show:

• How to create server-side monitoring configuration
• How to write alarm criteria, especially make it clear that only the metric, not the status, can trigger alerts.

each metric is output is identified by 2 variables %s.%s % (name,metric)

name is created by
name = instance.name.lower().replace('-', '_')

if name has spaces the metrics fail to be picked up by the alarm setup as spaces are not filtered

I ended up defaulting the name to a fixed string 'lb', as I am re using the alert syntax between load balancer checks.

For setting up checks I do not see the value of changing the name of the metrics based on the load balancer the name of the metrics should remain the same regardless of the load balancer.

There a various requests for pull data through SNMP. A plugin for advanced users to start using it and experiment would be a very good starting point.

Hey,

Please change line 75 to check for colons before doing the split operation.

Change as follows:

OLD:
 75         SLAVE_STATUS[i.split(':')[0].strip()] = i.split(':')[1].strip()

NEW:
 75         if ":" in i:
 76           SLAVE_STATUS[i.split(':')[0].strip()] = i.split(':')[1].strip()

This is to fix the error that came up when trying to parse the following section of the "show slave status" command:

      Replicate_Wild_Do_Table: 
  Replicate_Wild_Ignore_Table: 
                   Last_Errno: 1580
                   Last_Error: Error 'You cannot 'ALTER' a log table if logging is enabled' on query. Default database: 'mysql'. Query: 'ALTER TABLE slow_log
   MODIFY start_time TIMESTAMP NOT NULL,
   MODIFY user_host MEDIUMTEXT NOT NULL,
   MODIFY query_time TIME NOT NULL,
   MODIFY lock_time TIME NOT NULL,
   MODIFY rows_sent INTEGER NOT NULL,
   MODIFY rows_examined INTEGER NOT NULL,
   MODIFY db VARCHAR(512) NOT NULL,
   MODIFY last_insert_id INTEGER NOT NULL,
   MODIFY insert_id INTEGER NOT NULL,
   MODIFY server_id INTEGER UNSIGNED NOT NULL,
   MODIFY sql_text MEDIUMTEXT NOT NULL'
                 Skip_Counter: 0
          Exec_Master_Log_Pos: 32074
              Relay_Log_Space: 48474
              Until_Condition: None

The lines above that contain the rest of the ALTER query have no colons so the split() is failing and causing the program to crash.

Thanks!

Getting the following error:

Invalid metric line (line=metric accepted_conn uint32 ) - Metric line not in the following format: metric <name> <type> <value> [<unit>]

Looks like the values are not being printed correctly because if I run the script in command line I get the following :

status ok succeeded in obtaining metrics.
metric accepted_conn uint32
metric listen_queue uint32
metric max_listen_queue uint32
metric listen_queue_len uint32
metric idle_processes uint32
metric active_processes uint32
metric total_processes uint32
metric max_active_processes uint32
`metric max_children_reached uint32``

PHP 5.5.9-1ubuntu4.17
nginx/1.4.6