Monday, September 16, 2013

Parameter (Metrics) Thresholds: Do I still need to set them in PATROL?

On BPPM server, you can view all the data sent from each PATROL agent. You can set parameter (metrics) thresholds there including absolute thresholds such as 95% for file system utilization. Now you may wonder if you still need to set parameter thresholds in each PATROL agent.

In theory, if you send all PATROL data to BPPM server, it seems to be a good idea to have all parameter thresholds set on BPPM server only. Imagine how much time you can save for not having to set parameter thresholds in each PATROL agent and how much network bandwidth you can save for not having to send PATROL events to BPPM cells when those thresholds are violated.

In reality, the answer is yes. You still need to set parameter thresholds in each PATROL agent and let PATROL agent (not BPPM server) generate those events that violate absolute thresholds. You still need to send those PATROL events to BPPM cells. Let BPPM server generate intelligent events only and don't set absolute thresholds in BPPM server. The reason here is that not all PATROL data are in BPPM server.

First of all, PATROL agent does not buffer and resend data if it failed to send data to BPPM server for the first time. This situation can happen when there is a brief network outage such as a router is being rebooted. If it failed on the first try, data are lost forever. On the other hand, you may have already known that the connection between PATROL agent and BPPM cell is more robust as PATROL agent does buffer and resend events to BPPM cell with guaranteed delivery. BPPM server and BPPM cell were separately acquired by BMC Software from two different vendors. They use different communication protocols with different levels of robustness.

Second, PATROL agent only sends numerical data to BPPM server, not text data such as text parameter values and annotated data point values. Often those text data are needed as additional information for the events when numerical parameter thresholds are violated. For example, when using PATROL LOG KM, you may need to include information from a text parameter in the event to show the matched string. The only way to include information from a text parameter in events is to let PATROL agent (not BPPM server) generate the events. In addition, some PATROL KMs (e.g. LOG KM with custom events option, older version of Control-M KM, etc.) call event_trigger() to generate events without using parameter thresholds.

Last but not least, PATROL agent sends data to BPPM server every 5 minutes though PATROL agent may collect data more frequently. For example, CPU utilization is collected by PATROL agent every minute. This means that only every 5th value of CPU utilization is sent to BPPM server. Solely relying on absolute thresholds in BPPM server could result in delaying the alerts for 5 minutes or even missing the alerts altogether.

1 comment:

  1. how cloud i make the event that i generated by the function event_trigger() from a custom km to the Proactive Net Event List?

    ReplyDelete