[Beowulf] Big storage

Loic Tortay tortay at cc.in2p3.fr
Thu Sep 6 03:32:39 PDT 2007


According to Leif Nixon:
>
> > Even if we hadn't been close to a scheduled downtime, the applications
> > running on most of our X4500 are fault tolerant enough that we can
> > offline a machine for some time without a significant impact.
>
> These are dCache pools? If so, have you seen any dCache specific
> issues? You symlink the control directory to a non-ZFS filesystem?
>
Most of the X4500 we have are indeed dCache disk pools.

We had an issue with the (incredibly) long dCache startup time due to
some strange problem in Java/ZFS metadata interaction.

Our solution was indeed to move the dCache metadata to a UFS filesystem
on the mirrored system disks.

This issue and probably a few others were mentionned during the
X4500/ZFS related presentations during last HEPiX meeting in Hamburg.
I remember seing you there (or someone wearing a badge with your name
on it :-).

Specifically, Robert Petkus's presentation on disk systems tests (with
small input from me) and Martin Gasthuber's presentation on ZFS usage
at DESY and IN2P3 Computing Centre (for which I provided some data).


Loïc.
-- 
| Loïc Tortay <tortay at cc.in2p3.fr> -     IN2P3 Computing Centre     |



More information about the Beowulf mailing list