The Hibernate-based code has been merged. It recalculates category membership during each RTC update every 2 minutes so this bug has been resolved. Marking as fixed.

This should be fixed when I merge the DAO/JDBC-based RTC code.

I too am having issues with SLMs based on surveillance categories. For me, it seems to occur with nodes that have multiple interfaces when one of those interfaces has the same IP as a separate interface on a separate node.

For example, Switches with management interfaces 192.168.1.1 they rarely get hooked up in my environment. Any node with an interface with this same IP (regardless of it's state) with appear in SLM categories where any of their similarly addressed nodes appear legitimately.

Let me know if more information will be helpful. I'll be checking this thread.

I'm still seeing this problem as well. Using capsd, no provisioning at all (due to a legacy configuration that I haven't changed yet). Was running 1.8.10 on debian squeeze, but just upgraded to 1.8.12 today.

I have several different SLM categories, configured with different methods. Most of them use catincGroupName type definitions, but some use a 'nodeSysOID like ' definition.

In both cases, there are major discrepancies between what is shown on the home page SLM summary, and what is shown in the SLM pages.

EG, for one switch category (based on nodeSysOID), it shows "2 of 7" interfaces with outages, and 57.143% availability. The SLM page for that category shows only six switches in total, and only one has an outage. The one with the outage has 0% availability, which is correct, the remaining 5 each have 100%.

• node/interface totals are wrong

• average availability calculations are wrong (5/6 = 83%, and the 5/7 that the overview claims is still only 71% availability - there's no obvious way to get to 57% availability with the nodes I can see).

In a second case, we have a SLM category defined by membership of a specific surveillance category called "Broken_Switches", and uses catincBroken_Switches to define the SLM category. There are 4 switches in this category, and all are offline currently. The SLM overview shows outages on 4 of 4, which is correct, and shows a -50% availability (that's not a typo, it's showing negative 50%). Going to the SLM page lists only two switches.

I've just upgraded to 1.8.12-1, so everything has been restarted, and nothing changes here. I've also noticed the issue where changing surveillance group membership requires running rtc_restart in order to fix things, but there are these perpetually broken areas that no rtc_restart or opennms restart will fix.

This problem still appears in version 1.8.12-1. I have a node set with surveillance category "Pre-Production" and is displayed correctly on the home/SLM page but when I go move to the "Production" surveillance category the home/SLM pages do not update until I run the rtc_restart.sh or restart OpenNMS all together. I have all my nodes imported via a provisioning group where the surveillance categories are set then imported via provsiond, I have capsd disabled completely.

-Eric