MySQL Bugs: #46170: Log changes in MEM

Bug #46170	Log changes in MEM
Submitted:	14 Jul 2009 14:40
Reporter:	Bogdan Kecman	Email Updates:
Status:	Verified	Impact on me:	None
Category:	MySQL Enterprise Monitor	Severity:	S4 (Feature request)
Version:	2.1	OS:	Any
Assigned to:		CPU Architecture:	Any

Description:
It would be very useful if MEM could store changes in server details / topology and other variables it monitors .. stuff like:

server name
replication topology
server version
agent version
dash version
server configuration
...

so that informations like:
 - who made the change
 - when the change was made
 - what was the prior setup

can be retrieved from it.

How to repeat:
.

Suggested fix:
.

One additional set of things MEM could know about and store info 

 - new users added to MEM
 - new rules created / added / turned on

Thanks for creating the ticket. As a followup to what you've said I guess there are 2 different things:

1. recognise the changes and report them
2. record the changes so this can be monitored later and perhaps differences between a specific state and one or more earlier states can be viewed.

Probably first would be to recognise the change and report it.

Replication topology changes are not frequent but do happen. Merlin is currently configured to detect this and adjust the server groupings accordingly. However it does not send out a notification of the changes that it has detected. That alone would be useful.

Ideally that would be "removed x servers from group Y ", or "added x servers to new group Z". If this has to be split into a lower level initially that's fine.

In our case this would be summarised as:

- instance XXXX1 is no longer a slave of instance YYYYY1. It is a master so creating new server group GGGG1, and moving instance XXXX1 from group GGGG1 to GGGG2.
- instance XXXX2 moved from group GGGG1 to group GGGG2.

Providing a complete summary would be nicer, but the lower level topology changes would be fine.

We often change servers names as we prefer server123 to server123.aaa.bbb.example.com. With lots of servers server123 is nicer to work with. Notifying of a name change may be useful. It's done in the dashboard but if there's more than one DBA then the others aren't necessarily aware of these changes. This type of notification may not be appropriate to all sites (smaller ones with 1 dba/administrator may think that extra notifications are overkill), so please make this a toggable option somewhere.

Other states are recognised (server is unreachable) and notifiable, but please notify us when recovery takes place.... (otherwise we don't know).

In terms of logging the changes, that would be nice even if initially this is a log type table with just an event type, possible message, timestamp and possibly "who by" field. This may generate quite a lot of entries so make sure you clean up old entries periodically according to some strategy.

There's some overlap with this type of logging with the event notifications although it's assumed (falsely I believe) that a DBA will be there to recognise every single alert, something we don't do. These then pile up "untouched/unloved".

Just some extra ideas.

This is related to Bug #34237