Bug #53888 Suggested reorganization of the MEM connection graphs
Submitted: 21 May 2010 14:18 Modified: 9 Aug 2010 11:20
Reporter: Shawn Green Email Updates:
Status: Won't fix Impact on me:
Category:MySQL Enterprise Monitor: Web Severity:S4 (Feature request)
Version:any OS:Any
Assigned to: Andy Bang CPU Architecture:Any

[21 May 2010 14:18] Shawn Green
The existing 4 "connection" graphs are clunky and hard to use. Here is a suggestion on how to reorganize the same information, and a bit more, into two graphs. 

How to repeat:
use MEM

Suggested fix:
Currently there are 4 connection graphs in 2.2:
- Connections
- Connections aborted
- Connection cache
- Connections maximum

These are too "spread" about and need a bit of change:

I would suggest you change this to:

Connections (with the following fields)
- max_allowed
- thread_cache_size
- total
- running

Reason: this provides a much clearer indication of the state of the connections:
1. normally you might expect the total connections to be less than the thread_cache_size
2. seeing the running value and how it compares to either the maximum or the thread_cache size is also important

Having this information in one graph rather than the current 2 is I believe more helpful.

The second graph would be:

Connection activity:
- connections / minute
- connection threads created / minute
- client aborts / minute
- connect aborts / minute

Again these are all minutely "rates".
- we expect the connection threads created / minute normally to be 0 IFF the thread_cache_size is appropriately configured.
- both client or connect aborts are an indication of problems

Having these 2 graphs rather than 4 shows the same information and I believe is clearer and more appropriate than the current usage.
Please consider changing this as requested as I find the current layout rather fiddly and inconvenient to use.
[9 Aug 2010 11:20] Mark Leith
The Connection related variables were originally split out in to the current 4 graphs because the variables that are graphed can often be wildly different in terms of counts/sizes. That is, max_connections can often be set to 1000, whilst the number of running connections might be 8. This causes scale issues with the graph that make the smaller scale series much harder to read.