Hi Experts,
We have 2-node failover cluster in data center A for PROD. Recently we built another 2-node failover cluster in data center B for DR purpose. Our windows team mistakenly built DR nodes on a separate windows cluster, therefore we cannot make use of Always-On AG as it requires all nodes must belong to the same windows cluster. We are then implementing database mirroring between PROD and DR, but we now running into problem: We have 300 databases to mirror. While mirroring one database at a time and up to half way, we notice the current_workers_count had already exceeded max_workers_count. We have increased the "Maximum worker threads" from default 640 to 2000 on DR. So far, the PROD (12-core/264GB mem) and DR (12-core/264GB mem) still perform ok.
My questions are:
- How big is the risk of having "Maximum worker threads" way higher than the default value?
- If we remove database mirroring and rebuild the DR nodes to make them part of the same PROD windows cluster so we can make use Always-on AG, would we run into this "Maximum worker threads" issue again?
I need your expert advise to weigh my options:
- Continue database mirroring for the rest of databases and risking perf due to "Maximum worker threads"
- Get it of db mirroring, rebuild DR, make use of Always-On AG
- Leave DR the way they are, remove db mirroring and implement transactional replication
BTW, I don't have a test environment to test these options out.
Thanks much for any opinion
Thanh Nguyen
We have 2-node failover cluster in data center A for PROD. Recently we built another 2-node failover cluster in data center B for DR purpose. Our windows team mistakenly built DR nodes on a separate windows cluster, therefore we cannot make use of Always-On AG as it requires all nodes must belong to the same windows cluster. We are then implementing database mirroring between PROD and DR, but we now running into problem: We have 300 databases to mirror. While mirroring one database at a time and up to half way, we notice the current_workers_count had already exceeded max_workers_count. We have increased the "Maximum worker threads" from default 640 to 2000 on DR. So far, the PROD (12-core/264GB mem) and DR (12-core/264GB mem) still perform ok.
My questions are:
- How big is the risk of having "Maximum worker threads" way higher than the default value?
- If we remove database mirroring and rebuild the DR nodes to make them part of the same PROD windows cluster so we can make use Always-on AG, would we run into this "Maximum worker threads" issue again?
I need your expert advise to weigh my options:
- Continue database mirroring for the rest of databases and risking perf due to "Maximum worker threads"
- Get it of db mirroring, rebuild DR, make use of Always-On AG
- Leave DR the way they are, remove db mirroring and implement transactional replication
BTW, I don't have a test environment to test these options out.
Thanks much for any opinion
Thanh Nguyen
ThanhNguyen