[HCS-D] netapp notes

Matt Gline gline at fas.harvard.edu
Fri May 20 01:16:03 EDT 2005


Well Ladies and Gents, 

The time has come to make some important decisions. The netapp works
brilliantly. Currently, it's set up to use 7 active disks with one hot
spare, and the disk space is in two volumes - the root partition, 30
gigs large or so, which was on there when we got it and contains the
ONTAP operating system, and a big 300 gig volume (240 or so gigs usable
space plus 60 gigs reserved for snapshots). Both of these are mounted on
frog, so feel free to run df and ogle :) For those interested in boring
technical details, right now it's connected to frog directly (frog has
two ethernet interfaces, thanks to Yves and Matt Fasman who put the
second one in this morning). Bill Ouchark has suggested he'll be able to
find us a vlan capable fast switch - hopefully a gigabit switch - in
which case I think we'll set it up with two vlans - one external vlan,
one internal "storage" vlan. There are some complications involving the
speed of our link, but these are kinks to be worked out.

The questions I'd like to answer in short order, as I'd like to copy
/usr/home over very soon (potentially early next week) and start using
the thing, are:

1) Should we adjust our disk allocation scheme? Is there any reason we
want to break up the storage in a different way? We get to set different
snapshot rules on a per-volume basis, ONTAP's notion of dump seems
volume aware... is there any reason, for example, we'd like these things
to be different for people than for groups? If I understand flex volumes
correctly (there's an expert on this list who will I'm sure read this
carefully and correct all my mistakes) we can always resize things
later, so I don't think this decision is particularly pressing but it's
worth thinking about.

2) What do we want to do for a snapshot schedule? That is, when do we
want to take snapshots, and how long do we want to keep them for? We can
keep some number of weeklies, some number of nightlies, and some number
of hourlies, and can determine on which hours the hourlies are taken. I
think we probably want to store about 3 past weekly snapshots, because 3
or 4 weeks is the amount of time it takes some student groups to realize
something is missing or broken...

3) The Netapp supports some really cool things like WebDAV,
which it might be fun to take advantage of and offer to student groups
if we can figure out a paradigm for doing so. If we want to try this, I
think we'll have to scrap the whole "keeping the storage network
separate" paradigm. Thoughts?

I'm sure there's more I'm leaving out, but that's all I've got for the
moment. I'd also like to move our quota management over to the netapp,
which can handle user quotas although it needs first to have some
concept of user. It certainly supports NIS (ick) but I think if I
understand correctly it also supports LDAP, which is something we've
been wanting to set up anyway...

Cheers,
--Matt



More information about the hcs-discuss mailing list