Igor, The fix for this issue is merged to master. It will be included into Ignite 2.7. Until then you can use one of the nightly builds, available at https://ci.ignite.apache.org/viewLog.html?buildId=lastSuccessful&buildTypeId=Releases_NightlyRelease_RunApacheIgniteNightlyRelease&tab=artifacts&guest=1
Denis пн, 6 авг. 2018 г. в 15:01, Denis Mekhanikov <dmekhani...@gmail.com>: > Igor, > > There really is a memory leak in this place. > Thank you for the analysis! This is very cool! > > I filed a JIRA ticket for it: > https://issues.apache.org/jira/browse/IGNITE-9196 > Feel free to assign it to yourself and fix it, since you already debugged > through it. > If not, then I think, it will be fixed in Ignite 2.7 anyway. > > Denis > > сб, 4 авг. 2018 г. в 19:11, igor.tanackovic <igor.tanacko...@gmail.com>: > >> Hi, >> >> I want to discuss about potential memory leak in Ignite (I imply potential >> mem leak as I'm not 100% sure I do something wrong ;)) >> >> However, believe I found a corner case which triggers memory leak even >> with >> the lates stable version. In my case, MapQueryResults fills MapNodeResults >> map without being removed which leads to out of memory exception or huge >> gc >> major collections. I'm pretty sure the issue is in *fetchNextPage* method >> of >> MapQueryResult class as this method never returns *true* if you have >> result >> set with pageSize*n records. >> >> Actually, we had a major issue with memory in our production cluster and >> the >> only workaround was to limit all queries bellow page size (currently >> 1024). >> >> >> Regards, >> Igor >> >> >> >> >> -- >> Sent from: http://apache-ignite-users.70518.x6.nabble.com/ >> >
