>> One week ago we upgraded about 300 servers from 2.6.20.20 to 2.6.24.2.
>
> Painfull.
OK not really - i've tested the new kernel on all models. And it works
fine... the DN state only comes sometimes - absolutely not reproducable.
So my tests went OK and then we've done the update. I've now seen 5
servers in 7 days having this problem.
>> Now we get sometimes dozent of processes in state DN. The system is
>> completely idle - but the load is 20 or 90 or whatever. And it can
>> only be "repaired" by rebooting the system.
>
> please post sysrq-t (echo t > /proc/sysrq-trigger) output.
Sorry what key is sysrq? Can i also do this via SSH Console?
> Did you test with *one* server before upgrading all 300? If not, please
> do and try to upgrade in smaller steps, e.g. 2.6.20->2.6.21 and see when
> it breaks. Add a few *DEBUG .config options, they may help narrowing
> down the problem...
Problem here is that it is not reproducable...
Stefan
--
