[SunHELP] Latest Solaris 9 Recommended cluster issue

[Ahmed Ewing]

On Jul 19, 2006, at 2:30 PM, Connolly, Michael - EC Newton wrote:

> Just loaded Solaris 9 9/05 on a 280R. The install went fine. Now I've
> installed the latest Solaris 9 Recommended cluster (in single-user 
> mode, of
> course) and it has hosed the box!!!
>
> Errors seen:
> "undefined symbol 'ddi_can_recv_sig'
> "cannot load module ldterm"
> "cannot load module su"
>
> INIT: Command is respawning to rapidly"
>
> amongst others. As I've never had a cluster fail on this magnitude (I 
> cannot
> login) what is the best way to aproach this? Try to boot in single 
> user and
> back out the kernel patch (as this seems to be the root cause)?

I agree, it does look like something in the kernel got hosed, but 
honestly, this is a bit more common than you might think. One of the 
things many users fail to realize about the Solaris RPCs is that they 
only represent a list of all the latest versions of their patches; they 
are decidedly *NOT* tested together for functionality/compatibility. It 
is simply a grouping of the latest and greatest for your convenience, 
with nothing else guaranteed. As soon as any patch in the RPC list is 
updated, it is immediately reflected in the RPC. This is why RPCs have 
no versioning to speak of; they are a completely ephemeral construct in 
that regard, and fairly risky to use as blind upgrades on established 
production platforms. From Sun's README on all RPCs:

"A cluster grouping does not necessarily imply
that additional compatibility testing has occured since the individual
patches were released."

Contrast this with, say, the "Service Pack" for Windows operating 
systems, which represents a tested and proven-stable base level of 
patches... and even those have their issues :)

The last comparable offering from Sun would have been the Maintenance 
Update (MU), which brings an installation's patches (not features or HW 
support) up to that of a given install media kit's datecode; for 
instance, once upon a time, "Maintenance Update 36"/"MU36" might have 
brought any Sol9 installation up to the equivalent patch level of your 
9/05 disks, assuming 9/05 was the 36th media kit release since the 
initial Sol9 GA release. More info here:

http://www.sun.com/bigadmin/features/articles/solarisVersions.html

Unfortunately, MUs are no longer distributed for Solaris 9 as of 9/04:

http://www.sun.com/bigadmin/content/eof/eof.html

You may take some comfort in the fact that, since you did do a fresh 
install right before the attempted RPC installation, it should be 
fairly inconsequential to simply start again from scratch. Usually it's 
just a hiccup that causes corruption like that, but the trial-and-error 
process of pinpointing/fixing it piecemeal is often an exercise in 
futility.

Hope that helps...

-A