On Mon, Mar 11, 2013 at 7:25 AM, Iustin Pop <[email protected]> wrote: > On Mon, Mar 11, 2013 at 03:20:36PM +0100, Helga Velroyen wrote: >> Hi! >> >> On Mon, Mar 11, 2013 at 3:18 PM, Iustin Pop <[email protected]> wrote: >> >> > On Mon, Mar 11, 2013 at 06:28:26AM -0700, Guido Trotter wrote: >> > > On Mon, Mar 11, 2013 at 6:17 AM, Helga Velroyen <[email protected]> >> > wrote: >> > > > This patch updates the design doc "Design correct reporting of storage >> > > > free space". The modifications were chosen to not conflict with any >> > > > future changes of Ganeti regarding storage pools. >> > > > >> > > > Signed-off-by: Helga Velroyen <[email protected]> >> > > > --- >> > > > doc/design-storagespace.rst | 118 >> > +++++++++++++++++++++++++++++++------------- >> > > > 1 file changed, 83 insertions(+), 35 deletions(-) >> > > > >> > > > diff --git a/doc/design-storagespace.rst b/doc/design-storagespace.rst >> > > > index 919479f..67603e8 100644 >> > > > --- a/doc/design-storagespace.rst >> > > > +++ b/doc/design-storagespace.rst >> > > > @@ -20,21 +20,19 @@ interaction with different storage types. >> > > > Configuration changes >> > > > --------------------- >> > > > >> > > > -Each storage type will have a new "pools" parameter added (type list >> > of >> > > > -strings). This will be list of vgs for plain and drbd (note that we >> > make >> > > > -no distinction at this level between allowed vgs and metavgs), the >> > list >> > > > -of rados pools for rados, or the storage directory for file and >> > > > -sharedfile. The parameters already present in the cluster config >> > object >> > > > -will be moved to the storage parameters. >> > > > - >> > > > -Since currently file and sharedfile only support a single directory >> > this >> > > > -list will be limited to one. In the future if we'll have support for >> > > > -more directories, or for per-nodegroup directories this can be >> > changed. >> > > > - >> > > > -Note that these are just "mechanisms" parameters that define which >> > > > -storage pools the cluster can use. Further filtering about what's >> > > > -allowed can go in the ipolicy, but these changes are not covered in >> > this >> > > > -design doc. >> > > > +Add a new attribute "enabled_storage_methods" (type: list of strings) >> > to the >> > > > +cluster config which holds the types of storages, for example, >> > "plain", "drbd", >> > > > +or "ext". We consider the first one of the list as the default method. >> > > > + >> > > > +For file storage, we'll report the storage space on the file storage >> > dir, >> > > > +which is currently limited to one directory. In the future, if we'll >> > have >> > > > +support for more directories, or for per-nodegroup directories this >> > can be >> > > > +changed. >> > > > + >> > > > +Note that the abovementioned enabled_storage_methods are just >> > "mechanisms" >> > > > +parameters that define which storage methods the cluster can use. >> > Further >> > > > +filtering about what's allowed can go in the ipolicy, but these >> > changes are >> > > > +not covered in this design doc. >> > > > >> > > > Since the ipolicy currently has a list of enabled storage types, we'll >> > > > use that to decide which storage type is the default, and to >> > self-select >> > > > @@ -47,17 +45,23 @@ RPC changes >> > > > ----------- >> > > > >> > > > The noded RPC call that reports node storage space will be changed to >> > > > -accept a list of <method>,<key> string tuples. For each of them it >> > will >> > > > -report the free amount of space found on storage <key> as known by the >> > > > -requested method. Methods are for example ``lvm``, ``filesystem``, >> > > > -``rados``, and the key would be a volume group name in the case of >> > lvm, >> > > > -a directory name for the filesystem and a rados pool name, for >> > > > -rados_pool. >> > > > +accept a list of <method>,<key> string tuples. For each of them, it >> > will >> > > > +report the free amount of storage space found on storage <key> as >> > known >> > > > +by the requested storage type method. For example methods are ``lvm``, >> > > > +``filesystem``, or ``rados``, and the key would be a volume group >> > name, in >> > > > +the case of lvm, a directory name for the filesystem and a rados pool >> > name >> > > > +for rados storage. >> > > > + >> > > > +For now, we will implement only the storage reporting for non-shared >> > storage, >> > > > +that is ``filesystem`` and ``lvm``. For shared storage methods like >> > ``rados`` >> > > > +and ``ext`` we will not implement a free space calculation, because >> > it does >> > > > +not make sense to query each node for the free space of a commonly >> > used >> > > > +storage. >> > > > >> > > > Masterd will know (through a constant map) which storage type uses >> > which >> > > > method for storage calculation (i.e. ``plain`` and ``drbd`` use >> > ``lvm``, >> > > > -``file`` and ``sharedfile`` use ``filesystem``, etc) and query the one >> > > > -needed (or all of the needed ones). >> > > > +``file`` uses ``filesystem``, etc) and query the one needed (or all >> > of the >> > > > +needed ones). >> > > > >> > > > Note that for file and sharedfile the node knows which directories are >> > > > allowed and won't allow any other directory to be queried for security >> > > > @@ -73,18 +77,40 @@ Ganeti reporting >> > > > >> > > > ``gnt-node list`` will by default report information just about the >> > > > default storage type. It will be possible to add fields to ask about >> > > > -other ones, if they're enabled. >> > > > - >> > > > -``gnt-node info`` will report information about all enabled storage >> > > > -types, without querying them (just say which ones are supported >> > > > -according to the cluster configuration). >> > > > - >> > > > -``gnt-node list-storage`` will change to report information about all >> > > > -available storage pools in each storage type. An extra flag will be >> > > > -added to filter by storage pool name (alternatively we can implement >> > > > -this by allowing to query by a list of ``type:pool`` string tuples to >> > > > -have a more comprehensive filter). >> > > > - >> > > > +other ones, if they're enabled. Examples: >> > > > + >> > > > +> gnt-node list >> > > > +Node DTotal DFree MTotal MNode MFree Pinst Sinst >> > > > +mynode1 3.6T 3.6T 64.0G 1023M 62.2G 1 0 >> > > > +mynode2 3.6T 3.6T 64.0G 1023M 62.0G 2 1 >> > > > +mynode3 3.6T 3.6T 64.0G 1023M 62.3G 0 2 >> > > > + >> > > > +> gnt-node list -o dtotal/lvm/myvg,dfree/rados/myrados >> > > > +Node DTotal (Lvm, myvg) DFree (Rados, myrados) >> > > > +mynode1 3.6T - >> > > > +mynode2 3.6T - >> > > > + >> > > >> > > Could you add a note on how one is supposed to find about the >> > > "myrados" and "myvg" parts, and perhaps we can also have just >> > > dtotal/lvm that gives the "first/default" lvm. Also with storage >> > > pools, won't we go directly dtotal/pool_name without specifying "lvm" >> > > or "rados" in between? So perhaps we should "just" as an interim allow >> > > two levels, or allow three levels in between but go towards having >> > > two, in the future. >> > >> > Wait a second. I'm thoroughly confused now. >> > >> > I thought the discussion was that in the current stage we don't allow >> > custom names, but only one pool (key=type), and in the future key=type >> > migrates to key=name (pointing to the definitions)?! >> > >> >> That would drop the possibility to refer to vg and meta_vg seperately, at >> least in this interim state. Do we really want that? It is the only reason >> to have this three-part-definition as interim solution before storage pools >> are implemented. > > `metavg' should not be used for actual storage, so I don't see much gain > from having it available; htools doesn't take it into consideration, and > in general Ganeti presumes that the administrator has ensured there's > enough space (i.e. infinite) for it. > > So yes, I don't see a problem there. Handling the metavg correctly will > be a lot of work in any case… >
The other reason is that nowadays you can manually specify a vg, although then "unsupported things" happen, I guess (from allocation, etc). Since to fully support that we will need storage pools, I agree with this option, and to drop the three/level in favor of two, as long as we clearly specify that this "completely" supports only one vg/directory/etc and "manually specified"ones will need to be reconciled with pools later. thanks, Guido
