More Exchange stuff
Here's some more weirdness related to Exchange.
Basically we saw a lot of slowness, not all the time and not on all sites. This slowness would manifest itself in response times from the CAS/HUB server that sit behind a loadbalancer which comes from Kemp.
The CAS/HUB servers use the loadbalancer as the default gateway, now for all our traffic this worked extremely well, the only issue we got was this slow down. Even connecting to the server via RDP would be very slow and unresponsive, the cure was to vMotion the loadbalancer. More of a workaround than a fix but it helped.
After much troubleshooting and packet sniffing and a conversation with Kemp it was ascertained that if we change the NIC type from a VMXNET3 to a E1000 all our problems would disappear.
Surprisingly this actually worked.
Now the CAS/HUB servers are much more responsive and don't fall in to an apathetic well of despair.
Would I build an Exchange system that was fronted by a loadbalaner again? Probably not, most of our problems have ended up back at them, simple DNS round robin would have worked so much better.
Basically we saw a lot of slowness, not all the time and not on all sites. This slowness would manifest itself in response times from the CAS/HUB server that sit behind a loadbalancer which comes from Kemp.
The CAS/HUB servers use the loadbalancer as the default gateway, now for all our traffic this worked extremely well, the only issue we got was this slow down. Even connecting to the server via RDP would be very slow and unresponsive, the cure was to vMotion the loadbalancer. More of a workaround than a fix but it helped.
After much troubleshooting and packet sniffing and a conversation with Kemp it was ascertained that if we change the NIC type from a VMXNET3 to a E1000 all our problems would disappear.
Surprisingly this actually worked.
Now the CAS/HUB servers are much more responsive and don't fall in to an apathetic well of despair.
Would I build an Exchange system that was fronted by a loadbalaner again? Probably not, most of our problems have ended up back at them, simple DNS round robin would have worked so much better.
Comments
Post a Comment