So we have the following:
Backup Exec 2010 on a Windows 2008 x64 box Dell PE2950 with:
* Intel 10GbE SFF connected to 10GbE SAN (EQL PS6010XV)
* Dell SAS6 HBA connected to Dell TL2000 LTO5 2 head autoloader
When doing backups of the VMware machines using SAN option, we are getting less than optimal speeds when using 64K block/64K buffer. This is what we saw with the previous LTO4 (full height) TL2000 on a SAS5 HBA, connecting to the 10GbE SAN. When we changed the to 1MB block/1MB buffer, everything was great. We had our one job LTO4 getting 4593MB/min (and the LTO3 job getting 2497MB/min)
Changing to the LTO5 setup above, we get about 3200MB/min with the same job and the same VM's. Now this is with 64K block/64K buffer at defaults - which we expected to be slow, as that was slow with LTO4, so it's not really apples to apples.
But when we change to 1MB block/1MB buffer, the jobs don't verify, though they run very fast. They get to about 5GB of verifying, flying along at 10-15-20GB/min, then die, with "0xe00084c8 - the backup storage device has failed." Obviously, a backup that doesn't verify, isn't comforting. We've restored the tape on another drive, to another ESX server, and booted the VM up (network offline) and verified manually that the restore seems to work and the VM appears to be fine - but no one wants to sleep with verify's failing and/or no verify turned on.
We've tried talking to Symantec, and they are actively working with us. But I figured I'd post here and see if anyone else has had similar issues.
We are using:
* Latest LTO5 tape driver from Dell
* Latest LTO5 tape driver from Symantec (we've tried both. The Symantec one is dated 2008...?)
* Latest firmware on the TL2000
* Latest firmware on the LTO5 heads
* Latest firmware on the PE2950 doing the backups
* Latest firmware on the Dell SAS6 HBA
* Latest Dell driver for the SAS6 HBA
* Latest Intel 10GbE NIC driver
We're at a loss. The 64k/64k option on the LTO5 would be acceptable if it got the same'ish speed as the old LTO4 - but it doesn't come close, which is concerning.
Has anyone else seen anything similar?
