Hi Vic, > Bug ID 6566086 is fixed in 127111-08 according to the LDom 1.02 > release notes. This bug was about VDC I/O blocking and never failing > in a guest if it's service domain failed. So with that bug fix it > sounds like I/O will no longer block but will timeout at some point.
The default behavior hasn't changed and VDC I/O will still block. However each disk now has a timeout property which can be set to have the I/O timed out. When this property is not present or set to 0 then there's no timeout and I/O will not timeout. > Is the timeout tunable? Is it simply handled by sd_io_time now? You need a new version of the domain manager to set the timeout property for each disk. This version will be available with the next release of LDoms. In the meantime, you can use the vdc_timeout kernel tunable which will set the timeout for all disks of the guest domain. vdc_timeout is the number of seconds before the timeout is reached. If vdc_timeout is 0 then no timeout is set. To set the vdc_timeout, add this line in /etc/system on the guest domain and reboot the guest domain: set vdc:vdc_timeout = 30 In that case this sets the timeout to 30 seconds. > Also, that bug report said that timing out I/O was desirable in cases > where a ZFS mirror was used with multiple I/O domains for redundancy > in the guest. To what extent is ZFS supported in a guest with virtual > devices. When I've tried to create a zpool with VDC disks in a guest I > believe I ran into a problem where zpool complained that it couldn't > create the EFI label. Should that work? ZFS is supported with virtual disks. But if your virtual disk is a file or a volume then there is bug 6558966 (Virtual disks created from files do not support EFI labels) which prevents adding an entire disk to a ZFS pool. This bug is fixed in Nevada/Open Solaris build 82 and will be fixed in a patch after S10U5 is released. In the meantime, you will have a create a slice on your disk and add that slice to the ZFS pool instead of adding the entire disk, i.e. use: # zpool create test c0d1s0 instead of: # zpool create test c0d1 Rgds, alex.
