Issues
- NodeCategorySettingPolicy hit momentarily resolves open outagesNMS-6848Resolved issue: NMS-6848Benjamin Reed
- Allow provisioning policy to have a "PERSIST_ONLY" action to allow easier provisioning where only given IPs should be persistedNMS-5320
- Unable to use foreign source policy to set surveillance category by IP policyNMS-5319
- Provisiond temporarily deletes policy-based surveillance categories from existing nodes when synchronizingNMS-5059Resolved issue: NMS-5059Benjamin Reed
- Provisiond ignores foreign source category policies on subsequent synchronizationsNMS-4990Resolved issue: NMS-4990
- Policies in Provisiond return by default "false" if a SNMP attribute is nullNMS-4985
NodeCategorySettingPolicy hit momentarily resolves open outages
Description
Acceptance / Success Criteria
depends on
Lucidchart Diagrams
Activity
Benjamin ReedSeptember 25, 2014 at 7:50 AM
OK, this is fixed now. The event-handling for category and asset events was unconditionally removing outages, expecting them to be recreated on the next scan, I guess?
I modified the code to check each service individually to see if it still matched active filters, and add/remove from scanning as necessary.
Benjamin ReedSeptember 23, 2014 at 5:46 PM
OK, I had been doing this using the 'catinc' stuff and it appeared to be behaving properly. However, I just did it with a default poller configuration and I can confirm that this is still an issue. Provisiond is doing the right thing, but Pollerd or the outage service are not.
Benjamin ReedSeptember 22, 2014 at 10:07 PM
Provisiond has been fixed to not delete and add categories in phases, which repairs the issues with outages being cleared.
For details on the new design, see:
Jeff GehlbachSeptember 12, 2014 at 12:38 PMEdited
Very important note: when I've discussed this problem previously in conversation, I've assumed that reproducing it would require having a poller package whose filter keys on node category memberships, like:
<filter>catincSwitches</filter>
This turns out not to be the case. I was able to reproduce the problem using the stock poller-configuration.xml.
This issue created from report in support ticket https://mynms.opennms.com/Ticket/Display.html?id=3239
Steps to reproduce:
1. Create a requisition called "test". Create a node in that requisition called "test-sw-1". Add an interface to this node whose IP address is not reachable with snmp-primary=N. Add the ICMP service to that interface. Save the requisition.
2. Edit the foreign-source definition for "test". Add a policy called "Categorize", class "Set Node Category", key "category" => value "Switches". Match behavior = ALL_PARAMETERS. Add parameter key "label" => value "~.sw.".
3. Synchronize the "test" requisition. Wait for the test-sw-1 node to be created, populated, and to go into a nodeDown outage. Wait another moment for good measure.
4. Synchronize the "test" requisition again.
Expected result: nothing but maybe a nodeUpdated event on the test-sw-1 node
Actual result: new nodeDown outage created for the test-sw-1 node, and the previous outage was closed with no nodeUp event.
Additional remarks:
If I add a second node whose node label does not match the regex in the category-setting policy, that node's outages do not get summarily closed and re-created.