I'm new to SolarWinds, so please bear with me. I've recently installed SolarWinds 9.1 SP5 SLX and I can't get the polling engine to stay running during a network discovery import.
I'm utilizing existing hardware (at least for now). I originally built it on a Sun x2100 (1 - 2.2 GHz Dual Core AMD Opteron, 2GB RAM, 2-250GB SATA HDD - RAID 1 w/ separate OS and App partitions) running Windows Server 2003 R2 SP2 Standard and MS SQL 2005 Express. During a Network Discovery it would appear that the memory would get completely maxed out and the polling engine would stop. I scrounged up another 2 GB of RAM to bring the system to a total of 4GB, the memory didn't get maxed, but the polling engine still stopped, killing the import. After some initial attempts at troubleshooting including reinstalling I decided to throw more horsepower at the problem.
A clean install of Windows Server 2003 R2 SP2 Enterprise and MS SQL Server 2005 Standard on a Sun x4200 (2 - 2.2 GHz Dual Core AMD Opteron, 8GB RAM, 4 - 73GB SAS HDD - RAID 0 w/ separate OS and App partitions) later and I am still having issues with the polling Engine stopping during a discovery import. During the import the CPU usage typically stays around 25% occasionally jumping to 29%. I've never seen memory usage go above 2.26GB. The only way I've found to get the polling engine to restart is to kill all the frozen Network Discovery, restart the machine, and then wait for up to 30 minutes for the polling engine to start again. Web stays accessible and responsive throughout the polling engine stopping.
I've tracked how many nodes and interfaces were imported each time using the numbers in the NetPerfMonService.log and I have been getting fewer and fewer devices each time:
743 7942 (granted a lot of these were Cisco IP Phone without SNMP running)
I've tried tracking which devices it is hanging on, but if go back and attempt those devices again after a restart they import fine. I've tried breaking my imports down into smaller groups, but after a few small groups the same problem occurs.
I'll be happy to provide more data if anyone can suggest something, keep in mind that I have to work with the hardware that I have at least for now.