Hi Susan,

> With CUIR, CP tells the control unit it may not take the last path
offline.
> What happens when CUIR is disabled?  What happens if it is enabled, but
> z/VM does not support it?  Does the CE have to intervene to prevent the
> last path from being taken offline?

If CUIR is disabled, there is nothing for CP to react to.  I don't know
the
specifics, but it will probably require Operator intervention to VARY OFF
the affected path(s) manually in order to prevent I/O errors from
occurring
during the maintenance.  CUIR makes life a lot easier, and so if you have
CKD volumes I'd think you'd want to use it.

Again, z/VM does support CUIR, and you're going to have to go back quite a
ways to find a release that doesn't.  The caveat in this discussion is
that
we don't support CUIR for SCSI volumes, because such a thing does not
exist.

> One thing I did notice yesterday... while the EDEVs show multiple paths,
> the Linux guest does not appear to see both CHPIDs... using lscss the
edev
> only showed one CHPID.  Not sure if this is an issue with lscss not
> reporting the multiple paths correctly, or whether z/VM is not letting
the
> linux guest see the second path.

This is correct, and ties back to Alan's comments the other day.  For a
multipath EDEV, examine the output of QUERY EDEV xxxx DETAILS and QUERY
PATHS
xxxx.  The latter will only show one CHPID (that of an FCP subchannel used
within the EDEV) and reflects what Linux will end up seeing.  Since it is
possible for multiple paths to use the same FCP subchannel, but go to
different target WWPNs, we cannot rely on FCP CHPIDs being unique for each
path and thus cannot reflect one for each path defined in the EDEV.

This is the case for a lot of simulated/virtualized device types, so while
understandably confusing, isn't actually causing a problem.  Linux (or
other
guest) should be issuing I/O with a path mask of xFF, meaning any
available
path should be used for I/O, rather than relying on the CHPIDs it is
capable of seeing.

Hope this clears up your remaining questions.

Regards,
            Eric

Eric Farman
z/VM I/O Development
IBM Endicott, NY

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to