[Halld-offline] Request for more Lustre-based disk space

[Chip Watson]

Mark et al,

The new servers are 400 TB raw (40 10 TB drives).  After raid overhead 
and then allocating 80% they each deliver ~220 TB.

We are also retiring 4 2014 servers, 3 belonging to ENP, which is a loss 
to ENP of 224 TB (although it is not gone yet, just evaporating slowly 
at tens of TB per week perhaps).  So ENP's growth based on recent 
purchases is only 216 TB (>30%), with perhaps a higher bulge as the old 
nodes are retired.

Finally, ENP has to date been running with more than its fair share.

I will again tweak the pool sizes to track what is currently online.   
The sum of the guarantee in any pool needs to be less than 70% (still 
under consideration), but ENP's volatile guarantee is much too high, 
leading to a need to either reduce existing guarantees there or to 
increase the pool size there (meaning the /cache pool can't increase).

I do believe additional disk space is needed by ENP in FY2019, but we 
need to first analyze what we need, and whether the current file I/O 
pattern of GlueX can be improved.  If not we will have to buy SSD at 5x 
the cost / TB to address this problem.  I look forward to our scheduled 
discussion of hardware planning.

We can slow down the draining of the 2014 nodes a bit and maybe get 
additional capacity ordered before those are fully drained.

Bottom line: we can make tiny tweaks (and will within 24 hours) but 
probably below what you are requesting.

Chip



On 9/20/18 12:23 PM, Mark Ito wrote:
>
> Sandy,
>
> Now that the new Lustre servers have arrived (2 systems for physics, 
> 400 TB each), we would like to make a request for a near-term increase 
> in space. My understanding is that only one of the two systems is up 
> and running so more space will be available in the intermediate term 
> when the second system is installed.
>
> *Cache*
>
> Parameter
> 	Current Value (TB)
> 	Requested Value (TB)
> High
> 	450
> 	450
> Guarantee
> 	200
> 	300
> Pin Quota
> 	160
> 	250
>
> *Volatile*
>
> Parameter
> 	Current Value (TB)
> 	Requested Value (TB)
> High
> 	50
> 	200
> Guarantee
> 	30
> 	100
>
> Let me know if this is doable.
>
>   -- Mark
>
> -- 
> Mark Ito,marki at jlab.org, (757)269-5295
>
>
> _______________________________________________
> Halld-offline mailing list
> Halld-offline at jlab.org
> https://mailman.jlab.org/mailman/listinfo/halld-offline

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/halld-offline/attachments/20180920/7cacea30/attachment-0002.html>