7.4.2 Evaluating the performance of status polling

Evaluate the performance of status polling in your environment by using the information in the State Poller health check to quantify and assess the operation of the State Poller service.

State Poller health information tells you whether the Status Poller is able to keep up with polling requests.

Organization of this subsection

(1) Is the State Poller keeping up?

(1) Is the State Poller keeping up?

At any time, you can check the current health statistics about the State Poller service on the State Poller tab of the System Information window, as described in the following table.

Table 7‒1: State Poller health information
Information	Description
Status	Overall status of the State Poller service
Poll counters	Collections requested in last five minutes Collections completed in last five minutes Collections in process Collection request delays
Time to execute skips in last five minutes	The number of regularly scheduled polls that did not complete within the configured polling interval. A non-zero value indicates that the polling engine is not keeping up or that targets are being polled faster than they can respond. What to watch for: If this value continues to increase, there are problems communicating with the target or NNMi is overloaded. Action to take: Look in the `nnm.log` file for messages for the classes beginning with the string `com.hp.ov.nms.statepoller` to determine the targets for the skipped polls. If the skipped polls are for the same targets, change the configuration to poll these targets at a less frequent rate or to increase the timeout for these targets. If the skipped polls are for different targets, check the NNMi system performance, especially the available memory for `ovjboss`.
Stale collections in last five minutes	A stale collection is a collection that has not received a response from the polling engine for at least 10 minutes. A healthy system will never have any stale collections. What to watch for: If this value increases consistently, there is a problem with the polling engine. Action to take: Look in the `nnm.log` file for messages for the classes beginning with the string `com.hp.ov.nms.statepoller` to determine the targets for the stale collections. If the stale collections are for a single target, unmanage the target until you can resolve the problem. If the stale collections are for different targets, check the performance of the NNMi system and the NNMi database. Stop and restart NNMi.
Poller result queue length	What to watch for: Makes sure that this value is close to `0` most of the time. If it is greater than `0`, take the following action: Action to take: If this queue size is very large, `ovjboss` might be running out of memory.
State mapper queue duration	What to watch for: Make sure that this value is close to `0` most of the time. If it is greater than `0`, take the following action: Action to take: If this queue size is very large, check the performance of the NNMi system and the NNMi database.
State updater queue time	What to watch for: Make sure that this value is close to `0` most of the time. If it is greater than `0`, take the following action: Action to take: If this queue size is very large, check the performance of the NNMi system and the NNMi database.
State updater exceptions	What to watch for: This value should be `0`.

To Page Top