Hello esteemed fellows and keepers of the Patrol knowledge.
I've been struggling with my RT cloud, it's very distributed with a few pairs of RT servers in several locations all connecting up to a pair of RT servers that don't do anything but console server and BII4Patrol connections. Although I've increased the max connections to 600 it is no higher than that anywhere in the cloud. In total there are now 983 patrol agents, 3 console servers, and 12 BII4P adaptors. These are spread over 8 feeding RT servers each in pairs and the 2 top level RT servers. Many of the patrol agent are on one of the pairs which has 544 agents.
The problems I get are two fold.
- Firstly agents seem to drop off the cloud, although generally these loss of connections are fairly short in nature and I've learned to put up with them, but I suspect they are a symptom of something a miss.
- Second when a server reboots/crashes, and I suspect the patrol agent gets killed rather un-gracefully the console server seems to think the agent is still there for ages. I've watched for 10 minutes after a server has been rebooted and the patrol agent has remained 'connected' in the console. Eventualy the cServer shows the agent and disconnected and then reconnects it.
I suppose I'm wondering if people think that I have too many agents? I know that there is a limit of 700 connections for an RT server but I didn't think that this limited how many connections you can have in the cloud?
If I don't have too many agents, I suspect that I either have too many RT servers or too few? I used to have the connection timeouts much lower and this helped with closing connections when a server rebooted but caused more agents to loose connection here and there. Perhaps with more RT servers I can afford to decrease the timeouts again?
I'm afraid the RT servers are a mystery to me, it seems to come down to setting values and trying to figure out over time if that's made any difference. When I talk to support about the RT servers it seams to me they're doing the same thing. However the consequences of getting this stuff wrong in the cloud can be hideous.