Hi Ibrahim, I see you have 317 cache groups in your cluster `Full map updating for 317 groups performed in 105 ms.` Each cache group has own partition map and affinity map that require memory which resides in old-gen. During cache creation, a distributed PME happens and all partition and affinity maps are updated. This results in huge memory consumption and leads to long GC pauses.
It' recommended to use as few cache groups as possible. If your caches have the same affinity distribution you can place them at one cache group. It should help to reduce memory consumption. чт, 21 нояб. 2019 г. в 19:24, ihalilaltun <ibrahim.al...@segmentify.com>: > Hi Anton, > > Timeouts can be found at the logs that i shared; > > [query-#13207879][GridMapQueryExecutor] Failed to execute local query. > org.apache.ignite.cache.query.QueryCancelledException: The query was > cancelled while executing. > > huge loads on server nodes are monitored via zabbix agent; > <http://apache-ignite-users.70518.x6.nabble.com/file/t2515/SS_2.png> > > > just after cache creation we cannot return to requests, these metrics are > monitored via prometheus, here is the SS; > <http://apache-ignite-users.70518.x6.nabble.com/file/t2515/SS.png> > > for some reason, timeouts occur after cache proxy initializations (cache > creations) > > > > ----- > İbrahim Halil Altun > Senior Software Engineer @ Segmentify > -- > Sent from: http://apache-ignite-users.70518.x6.nabble.com/ >