sunit/swidth for HP P4500 Lefthand Networks storage arrays
The Linux host will be Debian Squeeze with a very recent BPO kernel, but running as a guest inside a VMware cluster. The storage platform is a dozen clustered HP P4500/Lefthand Networks "self load balancing" iSCSI arrays. The application is Dovecot IMAP using the mdbox mailbox storage format, which is basically a hybrid mbox/maildir+dedup format. Files will be 32-64MB or so, each containing multiple emails, with multiple files per mailbox directory. Mail indexes are internal to the files.
What is the best mkfs.xfs configuration for this scenario?
I'm guessing it would be best to simply use mostly, if not completely, the defaults, due to the way iSCSI packets are redirected on the fly to any storage node depending on load, by the Lefthand special sauce.
What about mount options?
Should I use barriers with the P4500s or disable them? TTBOMK the internal PCIe RAID controllers have BBWC, but the ~6GB of RAM on the P4500 mobos isn't battery backed, but for the typical external UPS. In this setup, from a physical hardware standpoint, iSCSI packets will be making at least 2 ethernet switch hops between the ESX nodes and the P4500s, with redundant links between everything, if that's a factor at all.
Signature · 6 lines
Stan _______________________________________________ xfs mailing list [email protected] http://oss.sgi.com/mailman/listinfo/xfs
Quoted 6 lines · On Donnerstag, 12. Januar 2012 Stan Hoeppner wrote:
What is the best mkfs.xfs configuration for this scenario?
I'm guessing it would be best to simply use mostly, if not completely, the defaults, due to the way iSCSI packets are redirected on the fly to any storage node depending on load, by the Lefthand special sauce.
I'd use defaults. We've recently switched to a NetApp storage, and with all the specialities it has also use the defaults.
Quoted 9 lines
What about mount options?
Should I use barriers with the P4500s or disable them? TTBOMK the internal PCIe RAID controllers have BBWC, but the ~6GB of RAM on the P4500 mobos isn't battery backed, but for the typical external UPS. In this setup, from a physical hardware standpoint, iSCSI packets will be making at least 2 ethernet switch hops between the ESX nodes and the P4500s, with redundant links between everything, if that's a factor at all.
Turn off barriers, I'd say. We use the NetApp over NFS (to VMware stores), and turned them off. I guess that's also correct to do.
As I understand them, barriers help to not loose blocks which the storage already received, so it doesn't matter how it's connected because the packets must have arrived there already. Can someone confirm?
Signature · 6 lines
mit freundlichen Grüssen, Michael Monnerie, Ing. BSc it-management Internet Services: Protéger http://proteger.at [gesprochen: Prot-e-schee] Tel: +43 660 / 415 6531
_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs