Gents, I created the ticket for custom executors: IGNITE-4699 [1] Alexander, I am not sure it makes sense since we are talking about distributed operations. CallerRunsPolicy is not suitable where user intentionally asks for job execution on remote node.
[1] https://issues.apache.org/jira/browse/IGNITE-4699 On Tue, Feb 14, 2017 at 11:22 AM, Alexander Fedotov < [email protected]> wrote: > Is it feasible to implement a saturation policy like CallerRunsPolicy? It > would allow for graceful degradation. > > On Mon, Feb 13, 2017 at 8:45 PM, Dmitriy Setrakyan <[email protected]> > wrote: > > > On Mon, Feb 13, 2017 at 8:18 AM, Alexey Goncharuk < > > [email protected]> wrote: > > > > > > > > > > We cannot compensate blocked threads with new ones forever. At some > > point > > > > we will have to limit new threads which will bring us to the same > > > problem, > > > > no? > > > > > > > > > > This means that a user recursively trying to call blocking tasks inside > > of > > > blocking tasks. We can as well detect this and fail with an analogue of > > > StackOverflowError (and emit a warning in the log in advance). > > > > > > > Got it. Agree. > > > > > > -- > Kind regards, > Alexander. >
