mon - Improvements

[Karl]

A few suggestions for improvements to the mon appliance:

- Option to convert any counter value to an average i.e. mySQL Selects - currently it shows the total number of selects that have occured, the option to have it convert it to an average based on the sampling period/pace would be much more useful in most circumstances so we'd end up with a value representing the number of selects/inserts/etc. per second over that period.

When a DB server has done 300m queries total in 3 months, the scale is going to be so small that you'd not notice a jump of 125k queries in a 5 minute period etc. wheras with an average, you're far more likely to notice this jump on the small graphs - Just as an example, a jump of 125k queries on a server that's done 300m already is < 1 pixel. Wheras that'd be a mighty big spike as an average (Normally 38 QPS, now 417 QPS).

- Option to combine two values in to one reading i.e. I've got 2 DB servers, I want to graph the total SQL SELECTS over both servers, so I want to say, "graph (DB1 Selects + DB2 Selects)" (again, optionally as an average over the pace time). Or I want to graph total available memory on my web servers etc.

- Option to copy a graph i.e. I have two DB servers, I setup a lot of graphs for DB1, that I want to change also do for DB2 - So I copy the graph for DB1, then edit the copy, and tell it to substitute DB1 for DB2 where there are performance counters that match i.e. Memory, CPU, SELECTS etc.

- Option to not re-arrange graph layout when browser window size changes - I've just spent 5 minutes arranging them how I want them, I don't want the browser window deciding it knows better and re-arranging and overlaping them :)

[PeterNic]

Karl,


A few suggestions for improvements to the mon appliance

Thanks for the suggestions -- all good ones. I have opened a single change request for now, SCR 2567, so that the changes can be considered as a batch. There is a possibility that we'll do an overhaul of the MON, to base it on an open source monitoring tool (e.g., RRDTOOL), now that browser Ajax support and graphing support has advanced a lot since the early days. An additional benefit would be that the MON appliance will be likely open source, so that anyone can make improvements (and share them if so desired).


Option to convert any counter value to an average i.e. mySQL Selects

As an aside, while I agree that you should be able to display different aspects of any counter (min/max/avg, etc.), I think that specifically MYSQL/MYSQLR should be fixed to provide the averages instead of the totals (which is the appropriate counter type for the current MON). It was an oversight when we added monitoring to the MySQL appliances. This is now tracked as defect SCR 2569.


Option to not re-arrange graph layout when browser window size changes

Whoa! I just tried to shrink the window. MON is not supposed to re-arrange graphs for you, ever. This is evil (and now known as defect SCR 2570 with high annoyance factor).

---

Another note: starting with AppLogic 2.3, the MON appliance has a web services interface for accessing all the monitoring data, as aggregated by MON. Some customers already use it to feed the data into things like Cacti and Nagios to monitor multiple apps. The same API can be used to provide some of the proposed capabilities prior to having MON support them.

--

Again, thanks for all the good feedback and suggestions. As a minor token of gratitude, please PM me your address and size, I'll be happy to have an AppLogic T-shirt sent out to you -- maybe we'll even do a limited series along the lines "I helped improve AppLogic" -- or, as digerata might say, "made it even more B.A." :). On the more serious side, we will look at all the proposed improvements when planning subsequent releases; fixes for the two defects I hope to be able to include in the first scheduled rev after the upcoming 2.4 release. If there is anything I can personally help you with, please let me know.

Best regards,
-- Peter

[Karl]

Probably out of the question here, but we found AnyChart.com fantastic for producing graphs/charts for our own internal systems.

Have you thought of exposing data, either from the appliances or from mon, via SNMP? As that's likely to be most compatible with everyone's monitoring systems.

Not had chance to try the web service yet for 2.3, it's on my list (a very long list).

[digerata]

Another note: starting with AppLogic 2.3, the MON appliance has a web services interface for accessing all the monitoring data, as aggregated by MON. Some customers already use it to feed the data into things like Cacti and Nagios to monitor multiple apps. The same API can be used to provide some of the proposed capabilities prior to having MON support them.


By chance is there any documentation laying around on how to hook cacti up to the monitor appliance?

I already have an appliance running cacti (branched from LUX5) and it's out terminal is connected to the aux terminal of my mon appliance. I can make queries from the cacti appliance to the mon appliance using the REST API (using wget, etc).

The part I'm missing is setting up cacti specific data inputs to read the REST output. Any doc pointers or tips would be greatly appreciated!

-Mike