SQL Clustering and Solarwinds High Availability Design Question

Currently working at a site where SQL clustering and Solarwinds HA are having issues. Both sites are connected via a MPLS WAN connection that is not a very high bandwidth connection (<100Mbps). There are two sites, each with a SQL Server, NPM server, and an Additional Polling Engine. The SQL servers are in a synchronous cluster where the standby will pick up in the event that the primary SQL server has issues. There is a Solarwinds HA Pool for the NPM servers and another HA pool for the additional polling engines. The problem is that when there is a SQL failover, the NPM server has to talk to the SQL server across the WAN which isn't a supported configuration as I understand the requirements. This also causes errors in the transaction log and the database and we end up having to restore the database. I'm not sure that this is the ideal deployment scenario so I'm for any feedback anyone may have on a more optimal design. I've included a diagram for reference. Thanks in advance for looking.

SolarwindsHA.PNG

Find more posts tagged with

Accepted answers

All comments

Seashore

Hi Kevin

I will assume you are using SQL Availability Groups as you are saying the SQL servers are in a synchronous cluster.

First, I would never set up a synchronous Availability Group,AG, with less than 1 GB low latency link between them. You will loose to much performance there. Maybe set them up in asynchrounus mode, with manual failover. I would after that manual failover restart orion services.

SQL over wan from Orion is not optimal but it works fine as long as you don't have to long latency there. Less than 100 ms I would say.

Failover with a synchrouns AG SHOULD not give any issues with transaction log. Probably you have other SQL issues due to high latency between the sites.

I would have the two SQL servers on the same site, on the same subnet.

nerengm

Make sure Solarwinds is pointing to the SQL Availability Group name, and not individual SQL servers. Also, your diagram implies that the active main server is in one site with the active SQL server, but the active APE is in the other site. That means the APE will have to connect to SQL over the WAN and likely to have performance issues. You'd better off keeping the APE active in the same site as the main server and SQL server, than polling over the WAN.

kevin.138

First, thanks for stopping by and commenting. This is exactly what is set up. We thought that it would be better to have the primary APE for site B to not have to poll across the WAN. But your saying it's better to poll across the WAN than to have SQL connections between the APE and SQL go across the WAN?