During inception review of this case, the ARC felt that
a spec was needed for FMA support (jdc-2) but a vote
could be held at a business meeting in lieu of a commitment
review. After investigation of the options, the project team
has proposed the following:

http://sac.eng/PSARC/2007/523/commit.materials/jdc-2.response

jdc-2   Will there be FMA ereport definitions?

Our original plan was to report fatal errors as ereports which will be
available via fmdump. I have attached the FMA portfolio we have been
working on for the same. During the internal reviews of the portfolio
it was mentioned to me that there is already work going on for
FMA for storage. The FMA case # for those projects are 2007/015 and 2007/016.
Since storage is used in a layer manner (raw disk, then filesystem
like zfs and then COMSTAR logical units over zfs), without this work
every layer will end up generating its own faults. This breaks the FMA
model where we want to generate only one fault per real failure.
After the FMA for storage work is done, every storage layer in solaris
will be able to consume it. But this also mean that we should defer the
FMA work for COMSTAR for now.

So the current plan then will be, to continue to handle errors internally
as we currently do and handle any COMSTAR relative offlines through the
stmf_ctl() interface which is part of the current PSARC. When the
backing-store diagnosis is available, we will have a follow-on phase to
deliver a COMSTAR 'agent' to subscribe to faults, and reflect their
effect relative to COMSTAR. At that time we will also be able
to introduce some simple COMSTAR-specific diagnosis relative to internal
state.
-------------

Updated materials are in the commit.materials 
<http://sac.eng/PSARC/2007/523/commit.materials/> directory.

Does the committee feel that a commitment review is required, or can we 
add this to
next week's ARC Business?

-- mark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: 
<http://mail.opensolaris.org/pipermail/opensolaris-arc/attachments/20080201/06700306/attachment.html>

Reply via email to