It explains a lot I thought that I am missing something in the docs ( So this is indeed what I see as it executed on the node where the services are running I think you call it coordinator. So this behavior is correct it's just need to be fixed in documentation ?
So is there a "safe"/recommended way to run a check/predicate on other remote nodes from the coordinator node in case of service deploy ? Instead of using closure as I have described. Thank You, Vladik On 5/28/18, 18:17, "aealexsandrov" <[email protected]> wrote: Hi, Indeed there is some unexpected behavior of the how it documented and how it works. I filed next issues related to service/cache deploying with node filters: https://issues.apache.org/jira/browse/IGNITE-8629 - There is no documentation about what on which node Ignite Predicate will be executed during service/cache deploying with NodeFilter https://issues.apache.org/jira/browse/IGNITE-8630 - Node filter IgnitePredicate executes twice during deploying on one single node cluster https://issues.apache.org/jira/browse/IGNITE-8631 - gniteInstanceResource doesn't inject the Ignite instance to IgnitePredicate during deploying cache with nodeFilter I propose you don't use the compute tasks in this predicate at the moment to avoid the problems with multiple executions of the apply method on the same node. BR, Andrei -- Sent from: http://apache-ignite-users.70518.x6.nabble.com/
