[evla-sw-discuss] CBE, lustre, mchammer, mctest down time today

[James Robnett]

   It turns out I'll need to reboot mctest as well.  I hadn't realized
it was running an completely unfriendly kernel.

   The upside is it will be a 1.8.4 lustre client afterwards.

James

James Robnett wrote:
> 
>   We're going to be re-addressing the Lustre install at the site
> today to accommodate routing between it and mchammer/mctest's secondary
> network interfaces.
> 
>   We will not be changing Lustre versions or kernel versions at this
> time.  We'll do that during some future update.
> 
>   I suspect the entire process to take about 30-45 minutes barring brain
> fade on my part.  I'd like to start around 9:30 or 10am,  likely right
> after the 9:00 meeting.
> 
>   The process is
> 
> 1) Shutdown CBE software (not computers) and unmount lustre
> 2) Stop telcal on mchammer/mctest and unmount lustre
> 3) Unmount lustre devices on MDS and OSS servers
>   3b) Change IP addrs on servers, disable mount on boot of lustre and
>       reboot (update DNS while they're booting)
>   3c) Do a write conf on OST's to update MDS info for them
>       tells storage server targets (OST) where the meta data server
>       (MDS) is now
>   3d) Testing mounting of MDT and OST partitions, reboot servers make
>       sure they work on boot
> 4) Mount Lustre on CBE, mchammer, mctest
> 5) Test test test
> 
>    As soon as I hear 1) and 2) are done I'll start on 3).
> 
> James
> 
>