Project

General

Profile

-1G - Client Security for CephFS


rturk    don't worry, nobody is in track 2 either :)  I think our lunch break was a bit short    12:29
rturk    most of the inktank devs in LA are still in the lunch room    12:29
rturk    I'll go get them    12:29
scuttlemonkey    hehe    12:29
scuttlemonkey    gregaf: we'll definitely want you for this one    12:31
rturk    he is on the way    12:31
gregaf    yeah, sorry, we were all shoveling food into our faces    12:31
gregaf    same hangout url?    12:31
scuttlemonkey    nah, ended up having to kill it...sent new    12:31
gregaf    ah, got it    12:31
*** sjustlaptop has joined #ceph-summit1    12:33
saras    boo    12:35
gregaf    yes    12:35
saras    at least i did not mess any thin    12:37
saras    some please define pg and osd    12:40
*** paravoid has joined #ceph-summit1    12:40
dmick    pg: placement group.  Sort of the first level of sharding objects across servers: grossly, an object hashes into a pg, and then the pg is placed    12:40
dmick    OSD: object storage daemon.  The intelligent storage entity that maintains replication and fault tolerance    12:41
dmick    and does most of the I/O    12:41
saras    dmick: thanks very much    12:41
rturk    (glossary coming in the docs very soon, I saw a sneek-preview)    12:41
*** cdl has quit IRC    12:42
*** cdl has joined #ceph-summit1    12:43
paravoid    our use case won't be covered by this    12:46
paravoid    the "perms per TB of data"    12:46
paravoid    that's a shame :/    12:46
saras    paravoid: what whould you need    12:46
paravoid    so basically we have a multi-tenant openstack nova setup and we'd like to give a shared filesystem per tenant, but over multiple VMs    12:47
Ryan_Lane    a similar use case to sharing directories via nfs    12:48
paravoid    hundreds of tenants (we call them "projects"), so a pool per project is excessive    12:48
paravoid    oh Ryan_Lane you're here    12:48
saras    paravoid: could join the hangout please    12:48
Ryan_Lane    where x set of instances can read, and another set can write, etc    12:49
Ryan_Lane    just think NFS exports :)    12:49
paravoid    Ryan_Lane: want to do that? I think you'll be much better at it    12:49
Ryan_Lane    hm    12:49
Ryan_Lane    I'm in a coworking space. i think I can, though    12:49
*** loicd has joined #ceph-summit1    12:50
Ryan_Lane    how do I join?    12:50
paravoid    I think rturk or scuttlemonkey have to invite you    12:50
paravoid    or maybe saras?    12:50
rturk    scuttlemonkey has a link    12:50
scuttlemonkey    yeah, sec    12:50
saras    not me here    12:50
scuttlemonkey    pm'd Ryan    12:51
saras    any one in the hangout could copy the link into the chat    12:51
saras    that works    12:51
scuttlemonkey    yep    12:51
Ryan_Lane    joined    12:51
paravoid    the discussion seems to be very technical now, maybe they'll leave some space for comments at the end    12:52
* Ryan_Lane nods    12:52
Ryan_Lane    so, I missed part of this    12:52
Ryan_Lane    what was their current idea on how to handle it?    12:53
paravoid    I missed half of it too, but there's this at the pad:     -- Permissions per TB of data would scale well, permissions per MB of data would be excessive    12:53
saras    soulds like their look at disc drive size    12:54
Ryan_Lane    I don't think our use case is abnormal    12:54
paravoid    Ryan_Lane is a colleague of mine    12:56
scuttlemonkey    are there specific questions now that we're in the last 5 mins    12:56
scuttlemonkey    ahh!    12:56
paravoid    we both work at wikimedia    12:56
scuttlemonkey    cool    12:56
gregaf    we have 4 minutes left, any questions we can address?    12:56
paravoid    and while I've been doing most of the ceph work so far, he's much better suited to talk about our cephfs plans    12:56
Ryan_Lane    I'm working with openstack nova, which is why we need the multi-tenancy :)    12:56
paravoid    we're expanding our use cases basically :)    12:56
AaronSchulz    paravoid: we have cephfs plans?    12:57
paravoid    AaronSchulz: labs    12:57
AaronSchulz    news to me :)    12:57
paravoid    replacement for gluster basically    12:57
paravoid    well, not formalized plans, it'd just be nice to reuse our expertise    12:58
saras    please    12:58