[Lustre-discuss] Nodes got evicting

[thhsieh]

********************************************************************
p.s.: Sorry that I should reply this Email to the lustre-discuss
      mailing list, but I uncarefully reply to the author who
      kindly replied my question yesterday. I should apologize
      to him that he may receive additional copy of this Email.
********************************************************************


Hello,

Thanks very much for your suggestion.

This morning our lustre filesystem crashed again. The same
head node was evicted. Hence I follow your suggestion to
upgrade it to version 1.6.6 (plus kernel 2.6.22.19). Now the
system is running and we are seeing what will happen ....

Could anyone remind me whether there are any known issues
of version 1.6.6 ?  So that I can make proper configurations
to avoid tentative problems.

BTW., do I have to disable the statahead in 1.6.6 ? If yes,
how to disable it ? Run the following command:

    echo 0 > /proc/fs/lustre/llite/*/statahead_max

in all the MGS/MDT/OST/client nodes ? Can I run that command
when the lustre filesystem is running ?

Thanks very much for your attention.


Best Regards,

T.H.Hsieh


On Tue, Dec 02, 2008 at 11:41:57AM -0500, Brian J. Murrell wrote:
> On Tue, 2008-12-02 at 15:18 +0800, thhsieh wrote:
> > Dear All,
> > 
> > We have a cluster (48 nodes) with Linux kernel 2.6.22.19 + Lustre 1.6.5.1
> > installed. After several weeks of testing run, we often encounter the
> > following error messages:
> 
> That's a pretty small snippet of messages, however...
> 
> > Please give me suggestions about how to solve this problem.
> 
> A good generic suggestion given that you are running 1.6.5.1 is to
> disable statahead.
> 
> You could also consider upgrading to 1.6.6 given that quite a number of
> issues were fixed with it.
> 
> b.
> 

> _______________________________________________
> Lustre-discuss mailing list
> Lustre-discuss at lists.lustre.org
> http://lists.lustre.org/mailman/listinfo/lustre-discuss