I just brought up a new Octopus cluster (because I want to run it on centos 7 for now) Everything looks fairly nice on the ceph side. Running FIO on a gateway pulls up some respectable IO/s on an rbd mapped image. I can use targetcli to iscsi share it out to a VMware node (cant use gwcli on centos 7) I can give a VM access to the iscsi target, as a "raw (virtualized) device)", and it still gives somewhat reasonable IO/s. 4,000/sec But as soon as I tell vmware to create a datastore on it, and carve a virtual disk out of it for the same VM to use... It chokes. 10 IO/s. yes, "ten per second". could anyone give me some suggestions on what to poke at? I have already tried simplifying the ALUA setup at the vmware CLI level. At the path level, I have tried setting VMW_PSP_FIXED_AP (VMWSATP_DEFAULT_AA) or Fixed(VMware) (VMSATP_DEFAULT_AA) but no luck. -- Philip Brown| Sr. Linux System Administrator | Medata, Inc. 5 Peters Canyon Rd Suite 250 Irvine CA 92606 Office 714.918.1310| Fax 714.918.1325 pbrown@xxxxxxxxxx| www.medata.com _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |