Please note that the issue only appears if backups > 0 for the cache. It
also doesn't matter if the cache is transactional or atomic.

On 2022/10/11 12:09:08 Anders Melgaard wrote:
> Hi all,
>
> I have a potential issue with continuous queries and memory consumption.
> I've run a test with a single cache with 3 CQs registered on one node.
> Another node joins and inserts one million entries and removes the same
> entries afterwards. This leaves the cache empty but it seems there's ~1,4
> million CacheContinuousQueryEntry objects left on the java heap and they
> won't be GC'ed. Inserting (and removing) another million entries causes
the
> number of entries to rise even further. Could this be a potential memory
> leak?
> I have tested this on another environment with 4 server nodes and ~20 CQs
> and here the CQ entries pile up much faster which eventually causing an
OOM
> node crash with more than 4,5 million CacheContinuousQueryEntry objects
per
> node hanging around on the heap.
> I have tested with versions 2.13 and 2.14.
> A simple proof of concept is available here:
> https://github.com/andersmelgaard/ignite-test2. Just execute the test
class
> twice.
>
> Best regards
> Anders
>

Reply via email to