Re: [Tarantool-patches] [PATCH rfc v2] memtx: fix out of memory handling for rtree

[Olga Arkhangelskaia <arkholga@tarantool.org>]

Hello Konstantin, thanks for the review.
It would be difficult to find right way without you.
I have spent more time looking at the code and I still think that

memtx_space_replace_no_keys

is the best place for the check.  First of all there is no 
memtx_rtree_build_next,
and secondly I am not sure that memtx_rtree_build_next should be 
implemented only with the check.


On 09/12/2019 17:07, Konstantin Osipov wrote:
> * Olga Arkhangelskaia <arkholga@tarantool.org> [19/12/09 16:49]:
>> When tarantool tries to recover rtree from a snapshot and memtx_memory value is
>> lower than it has been when the snapshot was created, server suffers from
>> segmentation fault. This happens because there is no out of memory error
>> handling in rtree lib. In another words, we do not check the result of
>> malloc operation.
>> To prevent this behaviour we simply reserve memory before replace operation for
>> rtree. And if there is not enough memory to be reserved - server will fail
>> gently with the "Failed to allocate" error message.
> It seems you're on track. You don't explain, however, why you had
> to add an additional reserve() which is on the side of the main
> execution flow (which is box_process1): during snapshot recovery
> the secondary keys are built in batches, not using box_process1,
> so the check sitting on the main execution track is not invoked.
>
> This begs the question: shouldn't you add the check to
> memtx_*_build_next() instead?
>
>