closing buffers used in flight producer (apache arrow flight - Java)

[a h <ah...@yandex.com>]

I have gone over multiple Java examples for apache arrow flight. In all the
flight producers that I have checked (for instance,
<https://arrow.apache.org/cookbook/java/flight.html)>, I cannot seem to see
any example (perhaps due to reading/testing incorrectly) where inside an
'acceptPut' method, the root or child allocator is being closed upon (I could
not locate a try with resources block). I realize a try with resources is
being used when the root allocator and cookbook producer is being created.
However, for a long running service, is that the desired setup or should one
use a try with resources within the acceptPut and similar methods to close the
child allocator (not the root one). I believe creating child allocators on the
fly might not be as fast but one could try to optimize that. I asked because
whenever I use the getHeadroom on the allocator after each acceptPut, I can
see that the headroom decreases without being reclaimed & the close method for
the auto closeable does not seem to be called. Quite happy to be shown what I
am doing wrong as my intent is only to improve the implementation I am working
on. Thank you.

Re: closing buffers used in flight producer (apache arrow flight - Java)

[a h <ah...@yandex.com>]

Actually ignore my previous question. I had made some changes and the previous
question is inaccurate.

22.05.2023, 10:09, "a h" <ah...@yandex.com>:

> I have gone over multiple Java examples for apache arrow flight. In all the
> flight producers that I have checked (for instance,
> <https://arrow.apache.org/cookbook/java/flight.html)>, I cannot seem to see
> any example (perhaps due to reading/testing incorrectly) where inside an
> 'acceptPut' method, the root or child allocator is being closed upon (I
> could not locate a try with resources block). I realize a try with resources
> is being used when the root allocator and cookbook producer is being
> created. However, for a long running service, is that the desired setup or
> should one use a try with resources within the acceptPut and similar methods
> to close the child allocator (not the root one). I believe creating child
> allocators on the fly might not be as fast but one could try to optimize
> that. I asked because whenever I use the getHeadroom on the allocator after
> each acceptPut, I can see that the headroom decreases without being
> reclaimed & the close method for the auto closeable does not seem to be
> called. Quite happy to be shown what I am doing wrong as my intent is only
> to improve the implementation I am working on. Thank you.