2011-01-26 11:28 AM
i do have some questions about an existing SME configuration.
- two mailbox-servers (dag) exchange 2010 hosted on two vSphere 4.1 VMs connected via FC to FAS3140 metrocluster
- snapdrive 6.3.p1
- sme 6.0p2
- each mailbox-server with 5 active dbs and 5 passive dbs separated with qtrees on one volume with corresponding logs in one lun on one volume
- each mailbox-server with 1 public-folder-db in a lun on one volume with corresponding logs in one lun on one volume
- each vm with 12 pRDMs created and connected with snapdrive "create disk"
- i tried to reorder these 12 pRDMs on different SCSI-Controllers, but after the next reboot no more disks in snapdrive available
- so, i learned that's not possible
- but, i learned too that while running a sme backup job with verify i was running out of scsi-ids because sme will force snapdrive to mount a snapshot and if the disconnect fails twice no more scsi-ids are available
- exchange logs are only truncated if the verify was successful?
- how do i connect sme to exchange?
on both mailbox-servers to each exchange-instance or on both servers to the dag-instance
- where do i create and run my backup jobs?
on both mailbox-servers with different times or on one server
with all databases or only with active or passive databases
- which backup-policies are you using (i know that's a difficult question)?
daily with 7 retention points and no up-the-minute (i saw fillung up the smeinfo with old logs)
weekly with 2 retention points and ...
- which volume-options are you using?
fractional reserve 0%
additional volume space 30%
snapshot autodelete forced by sme
how can i troubleshoot sme related vss writer errors in w2k8 and exchange?
questions over questions, but lets stop here its late.
2011-02-08 11:24 AM
i try to answer the your questions based on my expierence:
@ 1) pRDMs often mounted and dismounted - bad idea. Experience shows better use iSCSI with software initiator out of the vm to the target and let snapdrive connect the drive via iscsi. will be faster and much more reliable than pRDMs especially in the number of pRDMS you plan to use. depending on your hardware equipment you may need to add additional network cards for the iscsi connectivity.
@ 2) yes...
@ 3) depending on your needs. you can and should connect to the DAG an schedule your backup from there. the backups will be scheduled with a scheduled task on all members of the dag, but just one job will run. you can configure if the backup should run on the active / passive copy and if other copies should do a copy backup. this way backup is always done even if the dags failover or dag nodes are completly failing. the remaining node has the task and will run the job.
@ 4) pretty hard to answer.. really depends, protection manager integration available you may create weekly / monthly backups but usually customers have 7/14 daily backups in the default management group depending on sla. but nothing special.
@ 5) depends which errors you have. in my experience all issues with vss (especially those with unexpected provider error) are coming from not enough space on the volume due to fractional reserve 100% and more than 50% vol usage... you'll see the error imediatly on the file console during the backup run..
if you have any other questions, do not hesitate to ask...