On 06/21/10 08:27 AM, Dan Price wrote:
On Fri 18 Jun 2010 at 06:52PM, Cynthia McGuire wrote:

Template Version: @(#)sac_nextcase 1.70 03/30/10 SMI
This information is Copyright (c) 2010, Oracle and/or its affiliates. All 
rights reserved.
1. Introduction
     1.1. Project/Component Working Name:
         Solaris Instance UUID
     1.2. Name of Document Author/Supplier:
         Author:  Gavin Maltby
     1.3  Date of This Document:
        18 June, 2010
4. Technical Description
     See the case directory for more detail

This case might be easier to grok if it was called "crash dump UUID"
or "boot UUID" or some such-- "image UUID" seems vague, and this case
doesn't appear to be generating anything which is scoped beyond the use
case of crash dumps and live dumps (which seems perfectly reasonable).

I'm mostly interested in crash dumps, so I don't mind renaming it as
such.  There are times in service when it is valuable to ask a
customer to collect a bunch of info while the problem is experienced,
which might or might not included a forced crash dump or live
savecore.  For those purposes it can be useful to have label
under which such information can go, and that is what the
instance uuid can provide.  In cases where there have been
a few reboots/crashes it can be a real mess trying to
correlate which observations came from where.

I presume when you say "UUID" you are stating that the UUID will comply
with some particular standard?  RFC 4122?

Yes; thought I remembered adding that but it may be to a related
FMA portfolio.

My big question, thought, is "Why log it to syslog?"  The case provides
no rationale for this and seems to tempt customers to consume it--
especially as it marks the log message itself as "committed", which
seemed surprising to me.  There may be other committed syslog messages
but I don't know of them, and it seems like this would be something
to avoid (i.e. log messages are not an interface).

I have no need to log it to syslog for the immediate use in mind.
I figured that for the kind of service data collection scenarios
mentioned above that a record of the instance UUID would be
useful.  When you're performing some post-mortem panic analysis
it can be useful to correlate messages with the particular panic.
So I think that logging it has it uses.  I made it Committed
because its unlikely ever to have to evolve, and anyone
trawling it out of messages files can rely on its structure.
I was being generous there - I don't mind dropping the
stability level.

Thanks

Gavin

_______________________________________________
opensolaris-arc mailing list
[email protected]

Reply via email to