Dmitriy already created issue ( https://issues.apache.org/jira/browse/IGNITE-872). Valentin already found problem root (and added comment to issue). Sergi should look and fix.
On Fri, May 8, 2015 at 3:45 PM, Yakov Zhdanov <[email protected]> wrote: > Alex, can you please create a junit and ticket? > > --Yakov > > 2015-05-08 9:43 GMT+03:00 Alexey Kuznetsov <[email protected]>: > > > Checked with 2 nodes - scan query works fast. > > So, problem in case of single node. > > > > On Fri, May 8, 2015 at 8:17 AM, Alexey Kuznetsov < > [email protected]> > > wrote: > > > > > Val, yes I remember that I tested with one node. > > > But I will recheck with several nodes today and give more info. > > > > > > On Fri, May 8, 2015 at 2:38 AM, Valentin Kulichenko < > > > [email protected]> wrote: > > > > > >> Alexey, > > >> > > >> Do you have only one server node? In this case query is considered to > be > > >> local and pagination is disabled. Really looks like redundant > iteration. > > >> > > >> I also noticed that IgniteCacheProxy.query(Query filter, @Nullable > > >> ClusterGroup grp) method on line 347 ignores page size parameter > (added > > a > > >> comment in ticket). > > >> > > >> -- > > >> Val > > >> > > >> On Thu, May 7, 2015 at 8:32 AM, Alexey Kuznetsov < > > [email protected] > > >> > > > >> wrote: > > >> > > >> > Sergi. > > >> > > > >> > I put 1M entries in cache. Execute scan query from Visor and see 1M > > >> reads > > >> > in cache metrics (and I do not even hit "Next page" button). > > >> > Also more entries in cache results in more longer execution time of > > scan > > >> > query to wait for first page to show in Visor. > > >> > > > >> > Please take a look at VisorQueryJob code (line 75...). > > >> > > > >> > By default Visor use 100 rows as page size. > > >> > > > >> > I expect that take first 100 rows for scan query should be very > fast. > > >> > > > >> > On Thu, May 7, 2015 at 10:19 PM, Sergi Vladykin < > > >> [email protected]> > > >> > wrote: > > >> > > > >> > > Alexey, > > >> > > > > >> > > How did you come to this conclusion? I don't see anything like > that > > in > > >> > our > > >> > > code, we just take keySet or primaryKeySet iterator and fetch it > > page > > >> by > > >> > > page. > > >> > > > > >> > > Sergi > > >> > > > > >> > > 2015-05-07 9:58 GMT+03:00 Dmitriy Setrakyan < > [email protected] > > >: > > >> > > > > >> > > > I have filed a ticket for this issue: > > >> > > > https://issues.apache.org/jira/browse/IGNITE-872 > > >> > > > > > >> > > > Sergey, since it has to do with queries, can you please take a > > look? > > >> > > > > > >> > > > Thanks, > > >> > > > D. > > >> > > > > > >> > > > On Thu, May 7, 2015 at 1:47 AM, Alexey Kuznetsov < > > >> > > [email protected]> > > >> > > > wrote: > > >> > > > > > >> > > > > Hello All! > > >> > > > > > > >> > > > > Just found that if I have many entries in cache and do the > scan > > >> query > > >> > > > with > > >> > > > > page size 100 > > >> > > > > then my query seems to iterate over the whole cache and only > > after > > >> > > that I > > >> > > > > could fetch first page of data. > > >> > > > > > > >> > > > > I think that it should be reworked to return first page as > fast > > as > > >> > > > possible > > >> > > > > - without of waiting while scan query will be completed. > > >> > > > > > > >> > > > > Thoughts? > > >> > > > > > > >> > > > > -- > > >> > > > > Alexey Kuznetsov > > >> > > > > GridGain Systems > > >> > > > > www.gridgain.com > > >> > > > > > > >> > > > > > >> > > > > >> > > > >> > > > >> > > > >> > -- > > >> > Alexey Kuznetsov > > >> > GridGain Systems > > >> > www.gridgain.com > > >> > > > >> > > > > > > > > > > > > -- > > > Alexey Kuznetsov > > > GridGain Systems > > > www.gridgain.com > > > > > > > > > > > -- > > Alexey Kuznetsov > > GridGain Systems > > www.gridgain.com > > > -- Alexey Kuznetsov GridGain Systems www.gridgain.com
