Prachi, Mike, I appreciate your comments. I think it is a bug that there is a difference between filter logic of ClusterScopeStoragePoolAllocator and that of ZoneWideStoragePoolAllocator (ex. allocator skip iSCSI type storage pool or not when a volume is a RootDisk), I will unify by implementing it in AbstractStoragePoolAllocator#filter method.
It is new information for me that managed storages only support zone level. Thank you for let me know. But when iopscapacity parameter is not set to a primary storage (i.e. unmanaged storage), storageMgr.storagePoolHasEnoughIops always returns true. I suppose calling storageMgr.storagePoolHasEnoughIops from AbstractStoragePoolAllocator will not cause a problem, and it will simplify the code. I created a patch to explain my thought. Could you review it? https://reviews.apache.org/r/22717/ Regards, Noji 2014-06-17 19:19 GMT-06:00 Mike Tutkowski <mike.tutkow...@solidfire.com>: > I can comment on the IOPS part. > > The IOPS are currently only relevant when using managed storage (which is > currently only applicable at the zone level). That being the case, no > cluster (or local) storage takes advantage of counting IOPS that have been > handed out. > > Of course this logic might be expanded in the future, but that's how it > works today. > > > On Tue, Jun 17, 2014 at 2:47 PM, Prachi Damle <prachi.da...@citrix.com> > wrote: > >> >Why ZoneWideStoragePoolAllocator implements "filter" method (and doesn't >> call that of base class) rather than just using that of base class? >> >ZoneWideStoragePoolAllocator#filter method seems doesn't care "avoid" >> parameter, doesn't skip iSCSI type storage pool even if a volume is a >> RootDisk. >> >> The first question seems like a bug to me. >> >> I am not sure about the IOPS >> >> Prachi >> >> -----Original Message----- >> From: ynoj...@ynojima.net [mailto:ynoj...@ynojima.net] On Behalf Of >> Yoshikazu Nojima >> Sent: Tuesday, June 17, 2014 1:07 PM >> To: dev@cloudstack.apache.org >> Subject: ZoneWideStoragePoolAllocator#filter seems weird for me >> >> Hi, >> >> ZoneWideStoragePoolAllocator#filter seems weird for me. >> >> ZoneWideStoragePoolAllocator and ClusterScopeStoragePoolAllocator are >> allocator classes to select storage pool. >> They extends AbstractStoragePoolAllocator class, which provides "filter" >> method to exclude unavailable storage pools. >> >> Why ZoneWideStoragePoolAllocator implements "filter" method (and doesn't >> call that of base class) rather than just using that of base class? >> >> ZoneWideStoragePoolAllocator#filter method seems doesn't care "avoid" >> parameter, doesn't skip iSCSI type storage pool even if a volume is a >> RootDisk. >> (These functions are implemented in >> AbstractStoragePoolAllocator#filter method, which used by >> ClusterScopeStoragePoolAllocator.) >> On the other hand, AbstractStoragePoolAllocator#filter doesn't call >> storageMgr.storagePoolHasEnoughIops, so a cluster wide primary storage >> would be allocated more volumes than its designated IOPS capacity. >> >> Is there any difference between a zone wide primary storage and a cluster >> wide primary storage except its scope? >> If it is a bug, I'll fix it. >> >> Regards, >> Noji >> > > > > -- > *Mike Tutkowski* > *Senior CloudStack Developer, SolidFire Inc.* > e: mike.tutkow...@solidfire.com > o: 303.746.7302 > Advancing the way the world uses the cloud > <http://solidfire.com/solution/overview/?video=play>*™*
