Below are answers to questions asked during the recent webinar. If you have additional questions, please contact Janis Griffin at janisgriffin@confio.com
Q: When failing over with vMotion, does that restart SQL services like when failing over across the MS failover cluster manager (and in doing so disconnect/restart api connections to SQL)
A: vMotion leverages the complete virtualization of servers, storage and networking to move a running virtual machine from one physical server to another. This migration is performed with no impact to running workloads or connected users. During a vMotion, the active memory and execution state of the virtual machine is rapidly transmitted over the network to the new physical server, all while maintaining its network identity and connections.
Q: Windows2012 does know that it is virtualized…
A: See: http://partnerweb.vmware.com/GOSIG/Windows_Server_2012.html
Q: I had read that FT on VMWare has a limit on the resources available to the guest host (1cpu, etc.), is that true?
A: vCenter Server 4.x and vCenter Server 5.x support 1 virtual CPU per protected virtual machine.
Q: Note: the vSphere screens are showing vSphere 4.0. Current vSpher is 5.5 and has many perf upgrades and revised perf monitoring.
A: Yes, the screenshots were from version 4. There have been improvements in both the web browser and vSphere client in version5.
Q: Note: the vSphere screens are showing vSphere 4.0. Current vSpher is 5.5 and has many perf upgrades and revised perf monitoring.
A:Yes, the screenshots were from version 4. There have been improvements in both the web browser and vsphere client in version5.
Q: How can you tell if there is ballooning?
A: In the vSphere client, there are performance charts showing the rate of ballooning.
Q: What do you recommend for reservation RAM for SQL Server?
A: Depends on the workload in the SQL Server instance.
Q: Which ballooning perf counters to use?
A: Look in perfmon, for any VM counters, there you will see Memory Active in MB, Memory Ballooned in Mb, Memory Limit in MB…etc.
Q: Why would you set a limit that is lower than the memory configuration?
A: Limiting Memory is a way of keeping tighter control of the memory being used. If the configured amount is higher than the limit, the ESX can reuse the amount between the limit and config size for other VMs (ballooning). It can also use the same ballooning technique for this VM if it reaches the limit. However, if it is continually ballooning, you would want to increase the limit.
Q: Why would you have the capacity of RAM greater than the limit set for the VM?
A: See answer above.
Q: Is there a VMware or Microsoft recommended Layout for production and development SQL Server farm on a VMware cluster?
A: Check these out: http://www.vmware.com/files/pdf/sql_server_best_practices_guide.pdf http://technet.microsoft.com/en-us/sqlserver/bb671430.aspx
Q: How do I avoid "sleepy nic" issues?
A: See http://technet.microsoft.com/en-us/sqlserver/bb671430.aspx
Q: Do you have anything to add regarding the emergence of Flash Cache technologies?
A: VMware can use flash cache to help speed up read performance
Q: Does Ignite run on ESX host or virtual db server?
A: Confio can be set up to read from the ESX or Vcenter. It needs a read-only login and uses Vmware's API to capture the information already logged.
Q: Is it better to isolate your VM in a separate pool to give more resources?
A: Depends on the workload, if you have a highly transactional database that needs to perform. You may want to stand it up as the only VM or 2 on an ESX host. You would still get the benefits of Vmotion / DRS.
Q: Are standard SQL counters like buffer cache hit ratio reliable in VM environment?
A: Yes, the database counters will be correct as SQL Server knows how much memory and CPU it's consumed and the uses.
Q: What is average increase of performance does ignite cause for VM physical hosts?
A: Ignite puts very little load on the vCenter server as it only polls the log information every 20 seconds.
Q: FT supports more than 1 CPU after 5.5
A: VMware still states it only supports 1 CPU for FT at 5.5 http://www.vmware.com/pdf/vsphere5/r55/vsphere-55-configuration-maximums.pdf
Q: I have seen this message in our SQL Error Log "A significant part of sql server process memory has been page out." Could this be caused by memory pressure at the host level?
A: Yes, it could be at the host level but I would first check the VM and how it's configured with memory. Look for swapping at the Guest OS, then look at available memory at the host. If you have memory at the host, you can just reconfigure the VM with more memory.
Q: Is it better to have all databases on one dedicated datastore and physical ESX?
A: Depends on the database's workload & application flexibility requirements. If you need more flexibility or all databases are heavily I/O bound, you may want to separate on different VMs and/or hosts. You should always use SQL Server best practice by putting the log and data files on different devices.
Q: Do you find it useful to split tempdb, data and log as we do on physical machines?
A: Yes, you should still use those best practices on VMs.
Q: Is there any difference between thick provisioning and RAW disk?
A: Thick provisioning is allocating disk space to a VM before it's used. RDM in a VM environment is seen as a link to a raw device (there is no file system so no file system overhead). These are 2 different concepts.
Q: Should we limit the memory that SQL Server takes?
A: Yes, I believe you should set a max limit for SQL Server.
Q: Are there any VMware perfcounters base to use?
A: You can use PerfMon to view some of the VM performance counters.
Q: Do you think it is useful to have a min and max memory on SQL on VMware ?
A: Yes, I believe you should set a max limit for SQL Server and set your memory reservation to that value plus extra for OS and anything else running on the VM.
Q: Could you expound on the Ready metric?
A: It’s the amount of time a virtual machine waits in the queue in a ready-to-run state before it can be scheduled on a CPU. In other words, it's being starved for CPU as it waits.
