[OpenAFS] Re: Solaris 10 SPARC hang on shutdown

Patricia O'Reilly oreilly@qualcomm.com
Mon, 28 Feb 2011 14:02:14 -0800

I may butcher this a bit as I am not a kernel person. Reboot on Solaris will not work, period -- with or without the patch. This is not good for our environment as customers expect reboot to work the way it always has.

You can use the init scripts and all will be well unless you have an application that takes an inordinately long time to stop. We have a large automount map. Solaris goes through everything in the mnttab file umounting one at a time, serially. 

We want our customers to be able to happily reboot. The alternative is that they bug us about how afs sucks and is preventing their machines from shutting down.

Oracle gave our Solaris contact an appcert program to run on the afsd to see if it met their API specs on how a kernel app should behave. AFS failed that certification. Oracle holds AFS is not doing the right thing. Sun's reboot kills all user processes successfully except afsd. Oracle says that the afsd's are not user processes that never exit, they are really kernel threads and not user processes at all.

We need Sun's reboot to make an exception in AFSes case and stop waiting indefinitely for afsd to go down.


Derrick Brashear wrote:
> without looking I am betting those are slated for 1.4.15 which is probably about a month out
> Derrick
> On Feb 28, 2011, at 9:19 PM, Jeff Blaine <jblaine@kickflop.net> wrote:
>> On 2/28/2011 3:18 PM, Andrew Deason wrote:
>>> On Mon, 28 Feb 2011 12:10:54 -0800
>>> Patricia O'Reilly<oreilly@qualcomm.com>  wrote:
>>>> Even with the patch the wait is about an hour with the init script.
>>> To be clear, you mean it takes that long for all of the scripts to run,
>>> right? The OpenAFS script itself doesn't take an hour.
>> Patty,
>> FWIW, I applied the patches just now to 1.4.14 and
>> shutdown -g0 -y -i6 works properly for us (comes down
>> properly within 1 minute).
>> Devs: What's the timeframe to see these patches in an
>> official 1.4.x release?  Any idea?
>> Thanks again.
>> _______________________________________________
>> OpenAFS-info mailing list
>> OpenAFS-info@openafs.org
>> https://lists.openafs.org/mailman/listinfo/openafs-info
> _______________________________________________
> OpenAFS-info mailing list
> OpenAFS-info@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-info