MySQL Bugs: #43873: 'Slave I/O Thread Not Running' is always notified as 'Unknown'.

Bug #43873	'Slave I/O Thread Not Running' is always notified as 'Unknown'.
Submitted:	26 Mar 2009 3:58	Modified:	24 Jan 2013 23:51
Reporter:	Meiji KIMURA	Email Updates:
Status:	Closed	Impact on me:	None
Category:	MySQL Enterprise Monitor: Documentation	Severity:	S2 (Serious)
Version:	2.0.5	OS:	Any
Assigned to:	Assigned Account	CPU Architecture:	Any

Description:
'Slave I/O Thread not Running' is always notified as 'Unknown'.

MySQL Server: 4.0.26
MEM Server: 2.0.5.7153
MEM Agent: 2.0.5.7153

I build the simple Master/Slave replication using MySQL Server 4.0.26. After installing MEM and agents, I apply the replication rule (total 19) to Master/Slave with default settings.

At first, 5-rule are notified as 'Unknown'.

unknown  	unknown  	MeijiK-PC:3311  	Replication  	Slave Without REPLICATION SLAVE Accounts  	Mar 26, 2009 

1) Slave I/O Thread Not Running
2) Slave Has Stopped Replicating
3) Slave Has Experienced A Replication Error
4) Slave Execution Position Too Far Behind Read Position
5) Slave Without REPLICATION SLAVE Accounts 

I check them specifying 'All' for Severity.
Basically above 5-rule are always 'Unknown'. But 5) became 'OK' after several minutes.

After 1-hour the rest of them remains 'Unknown'.

How to repeat:
See Description.

Suggested fix:
[Workaround]

None.

[Suggested fix]
I check the agent of slave with log-level debug. it seems an agent send proper information of slaves, but MEM seems not to deal with them properly.

'Unknown' items on dash board.

Attachment: dash_board.JPG (image/jpeg, text), 105.21 KiB.

This is basically working "as designed". However, it could be documented for clarity. This issue could affect any rule that is relying on master_host and another data item (which is the case for many Replication rules) and will be most noticeable on the rules scheduled using the default of 6 hours. Basically it has to do with the evaluation alignment of master_host and the first time evaluation of the other data item. It will be resolved (not Unknown) the next time the other data item is evaluated (in 6 hours). To see the Unknown "go away" sooner, you could schedule the rule(s) for much less than 6 hours.

Bill,

I don't agree with your comment. I listed rules with default interval.

1) Slave I/O Thread Not Running(00:01)
2) Slave Has Stopped Replicating(00:01)
3) Slave Has Experienced A Replication Error(00:05)
4) Slave Execution Position Too Far Behind Read Position(00:01)
5) Slave Without REPLICATION SLAVE Accounts (06:00)

Only 5) is set with 6-hour, the rest of them is scheduled 1-minute or 5-minute.
The events of them should be modified in 5-minutes, so I think it is not a behavior expected.

please reproduce it with the agent logs set to debug log level, and attach to this bug.

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".