Forum Discussion
Jeremy_Bridges_
Nimbostratus
NimbostratusSCOM Pool or Pool Member Performance Indicators
From looking through the documentation, I know that there are a set of statistics available through the F5 management pack for pools or pool members:
Current Connections
Maximum Connections
Packets In
Packets Out
Total Bytes In
Total Bytes Out
Total Connections
First, a question about the Current Connections stat: does this number just reflect what were the open connections at the precise moment the counter is queried?
About the Maximum Connections counter, this is the maximum concurrent connection count over some time period, right? If so, how long a time period does the F5 use? Should this be better described as "Maximum Concurrent Connections during the last X minutes"?
Finally, does the Total Connections counter just increase idenfinately, or does it reset at regular intervals?
4 Replies
- So the Current Connections does correspond to a specific moment in time when the statistic is collected. In Operations Manager, you can check the timestamp if you are curious as to the exact time this was collected. You may also turn up the collection interval to 30 seconds instead of 5 or 10 minutes, as this will give you closer to "real time" current connections.
The Maximum Connections and Total Connections are statistics taken directly from the Big-IP as well. These are only reset by the user directly from the Big-IP itself. I think these do not get reset automatically by the BigIP and must manually be reset to 0, or a rule set up to reset them after a specific period of time has passed, although I'm not entirely sure how these counters work. 
Jeremy_Bridges_It would be nice to know how the Total and Maximum Connections stats work. The management pack documentation states that they are "avg/sec":
http://devcentral.f5.com/wiki/default.aspx/MgmtPack/BigIPHealthModel.html
But, that still is not entirely clear to me. How does one calculate the "average per second" for maximum or total connection counts?- So the way we calculate this avg/sec is by keeping track of the previous value received from the Big-IP as well as the timestamp. When the next value comes in, we take the difference of the 2 values and divide it by the amount of time that passed. For instance:
Received Bytes Out: 5302 @ 10:12:56
Received Bytes Out: 5518 @ 10:13:23
Works out as: 216 bytes / 27 seconds = 8bytes/sec which can be read as "the average increase in bytes out is 8 bytes per second".
That negative value you are seeing is actually a product bug that I was able to find today based on your logs. It looks like some packets are received in the wrong order, which creates a negative time span, which unchecked, then creates a negative average value. I've fixed the issue and it will be in the September release. - Jeremy_Bridges_Is the bytes out counter a "total bytes out" counter? In other words, if only one byte was received between each sampling would the counter look like this?
5021 @ 10:14:21
5022 @ 10:15:45
5023 @ 10:16:11
5024 @ 10:17:36
Or would it look like this?
1 @ 10:14:21
1 @ 10:15:45
1 @ 10:16:11
1 @ 10:17:36
