[Gluster-Maintainers] [Gluster-devel] Release 3.10 feature proposal: gfapi fix memory leak during graph switch

Rajesh Joseph rjoseph at redhat.com
Wed Dec 14 05:08:05 UTC 2016


On Mon, Dec 12, 2016 at 11:42 PM, Shyam <srangana at redhat.com> wrote:
> On 12/09/2016 08:06 AM, Rajesh Joseph wrote:
>>
>> During graph switch we do not perform much cleanup operation on the
>> old graph, leading to memory leak.
>>
>> + Inode table of old graph needs cleanup.
>>        - Fix inode leaks
>>        - Fix forget of each xl to free inode ctx properly
>> + The xl objects itself (xlator_t)
>> + The mem_accnt structure in every xl object.
>>       - Fix all the leaks so that the ref count of mem_accnt structure is
>> 0
>> + Implement fini() in every xlator
>>
>> Would need support from all other xlator owner to bring down memory
>> leak in each xlator.
>
>
> Rajesh, I have marked this for 3.10 tentatively. As you call out a per
> xlator task, could you elaborate what is needed from others?


Basically we need proper fini implementation in all xlators. That means all
xlators must free up all resources allocated from init to fini.

>
> We need to possibly get ack's from the various component maintainers so that
> this can happen for 3.10. So if you could share a list of things that need
> to be done/checked for each component and drive the same, in terms of
> timelines, we could get this feature complete.
>

We have most of the big ticket items but each xlators must come up with their
list of leaks. May be for this release we can target only most frequently used
xlators. I will update the list on the github project planning card.

Thanks & Regards,
Rajesh


More information about the maintainers mailing list