It is just flags / configuration. The APIs would just let you set the flags/configuration. You could also use APIs to read the offheap statistics. Basically just allocate the offheap memory using "off-heap-memory-size" and then configure one or more regions with offheap set to true. Any value stored in that region will be stored in offheap memory instead of heap.
On Wed, Jan 4, 2017 at 11:29 PM, Dor Ben Dov <[email protected]> wrote: > Kirk & Darrel & Hitesh, > > Thank you all for the detailed info on the off heap option. > > One more question, do I have any kind of API or it is purely flags / > configuration? > > Regards, > Dor > > -----Original Message----- > From: Kirk Lund [mailto:[email protected]] > Sent: יום ג 03 ינואר 2017 22:47 > To: geode <[email protected]> > Subject: Re: Off-Heap Annotations > > Oh yeah! We were hoping to get more feedback from users to find out what > they want or need for migrating a region from heap to off-heap. Right now > it doesn't support anything like that but it was definitely discussed. > > > On Tue, Jan 3, 2017 at 12:27 PM, Hitesh Khamesra < > [email protected]> wrote: > > > I was considering the case when one want to change its "value" through > > rolling upgrade. In an ideal world, I would expect that every node > > will have the same heap/off configuration, but other can happen. > > > > > > > > From: Kirk Lund <[email protected]> > > To: geode <[email protected]> > > Sent: Tuesday, January 3, 2017 12:16 PM > > Subject: Re: Off-Heap Annotations > > > > The region is required to be off-heap in every member in the cluster. > > Imagine memberA with 12 GB heap, 100 GB off-heap and memberB with 8 GB > > heap, 0 GB off-heap. What's the point of having replicated regionA be > > off-heap in memberA but on heap in memberB? The region will always be > > limited by the heap size of memberB. More importantly, the > > ResourceManager would not handle things correctly unless regionA is > > consistently off-heap in every member. The documentation also > > recommends that every member have the same off-heap-memory-size > > because of the way the ResourceManager is implemented as well. Any > > major differences in memory size between members is a good way to > > cause one or more of them to run out of memory unless you constrain the > amount of data that can go into the region in some way. > > > > > > On Tue, Jan 3, 2017 at 11:51 AM, Hitesh Khamesra < > > [email protected]> wrote: > > > > > >>>The first cluster member to create the region > > > wins and all the other members need to conform to the first one's > > > configuration. > > > Is there any technical reason to check this? Or we just want same > > > configuration across cluster. > > > > > > From: Darrel Schneider <[email protected]> > > > To: [email protected] > > > Sent: Tuesday, January 3, 2017 11:33 AM > > > Subject: Re: Off-Heap Annotations > > > > > > When a region is created you need to mark it as off-heap or on-heap. > > > Once it is created you can not change it on that region. > > > You can create and destroy regions on the fly. > > > > > > c > > > > > > On Tue, Jan 3, 2017 at 8:01 AM, Dor Ben Dov <[email protected]> > > > wrote: > > > > > > > Udo, > > > > > > > > This means it can't be changed on the fly in runtime ? Need to be > > flagged > > > > before? > > > > > > > > Dor > > > > > > > > -----Original Message----- > > > > From: Udo Kohlmeyer [mailto:[email protected]] > > > > Sent: יום ג 03 ינואר 2017 17:32 > > > > To: [email protected] > > > > Subject: Re: Off-Heap Annotations > > > > > > > > Hi there Gal, > > > > > > > > That is not possible. A region is either on-heap or off-heap. > > > > > > > > --Udo > > > > > > > > > > > > On 1/2/17 13:50, Gal Palmery wrote: > > > > > Ok, > > > > > So if I want to have just a certain part of the region off heap > > > > > and > > the > > > > rest of it on heap - how do I do that? > > > > > > > > > > > > > > > > > > > > -----Original Message----- > > > > > From: Dan Smith [mailto:[email protected]] > > > > > Sent: Monday, January 02, 2017 19:46 > > > > > To: [email protected] > > > > > Subject: Re: Off-Heap Annotations > > > > > > > > > > Hi Gal, > > > > > > > > > > The way to control what is on or off heap is when you configure > > > > > a > > > region. > > > > > Regions that are configured with off-heap=true will store all of > > > > > the > > > > values in off heap memory, regions with off-heap= false will store > > > > the values on the heap. > > > > > > > > > > -Dan > > > > > > > > > > On Mon, Jan 2, 2017 at 5:57 AM, Gal Palmery > > > > > <[email protected]> > > > > wrote: > > > > > > > > > >> Thanks Kirk. > > > > >> > > > > >> for example, before I call put, I'd like to indicate to geode > > > > >> server that I want to keep specific data off heap. how can I do > that? > > > > >> is there an api that will move data off or on heap? > > > > >> > > > > >> Gal > > > > >> > > > > >> -----Original Message----- > > > > >> From: Kirk Lund [mailto:[email protected]] > > > > >> Sent: Tuesday, December 27, 2016 06:33 > > > > >> To: [email protected] > > > > >> Subject: Re: Off-Heap Annotations > > > > >> > > > > >> Those annotations are part of the internal implementation of > Geode. > > > > >> The developers used it as a way to help keep track of > > > > >> references to > > > > off-heap. > > > > >> > > > > >> -Kirk > > > > >> > > > > >> > > > > >> On Mon, Dec 26, 2016 at 1:34 AM, Gal Palmery < > > [email protected]> > > > > >> wrote: > > > > >> > > > > >>> Hi everyone, > > > > >>> > > > > >>> Is this feature: > > > > >>> https://cwiki.apache.org/confluence/display/GEODE/Off- > > Heap+Annotatio > > > > >>> ns a part of the last release? (1.0.0-incubating) > > > > >>> > > > > >>> thanks, > > > > >>> Gal > > > > >>> > > > > >>> This message and the information contained herein is > > > > >>> proprietary > > and > > > > >>> confidential and subject to the Amdocs policy statement, you > > > > >>> may review at http://www.amdocs.com/email_disclaimer.asp > > > > >>> > > > > >> This message and the information contained herein is > > > > >> proprietary and confidential and subject to the Amdocs policy > > > > >> statement, > > > > >> > > > > >> you may review at http://www.amdocs.com/email_disclaimer.asp > > > > >> > > > > > This message and the information contained herein is proprietary > > > > > and confidential and subject to the Amdocs policy statement, > > > > > > > > > > you may review at http://www.amdocs.com/email_disclaimer.asp > > > > > > > > This message and the information contained herein is proprietary > > > > and confidential and subject to the Amdocs policy statement, > > > > > > > > you may review at http://www.amdocs.com/email_disclaimer.asp > > > > > > > > > > > > > > > > > > > > This message and the information contained herein is proprietary and > confidential and subject to the Amdocs policy statement, > > you may review at http://www.amdocs.com/email_disclaimer.asp >
