On 26 Oct 2023 at 19:26:15, Tim Allison <[email protected]> wrote:

>
>
> Request limits can fairly easily be implemented outside of Tika, but
>> resource isolation is not, so having a solution for that as well would be
>> very nice.
>>
> Isolation as in pipes? One file per forked jvm at a time?
>

Yes. I am assuming that most of the memory will be shared so the memory
overhead is not that high?



>
>> This SO (https://stackoverflow.com/questions/52495429/setting-xxmaxram)
>>>
>>> suggests that maxRAM is only a hint, not a strict enforcement. 😟
>>>
>>>
>> That is my experience as well.
>>
> +1 :(
>
>> Should we allow users to specify a ulimit on the forked process?
>>>
>>>
>> In my limited testing ulimit -m is also ignored, at least when running in
>> a container.
>>
> Argh.  I guess we could have the monitoring process monitor Docker
> containers instead of forked processes? :P
>
>>  I haven't had a chance to figure out if we can do this more easily at
>> the cxf level.
>>
> This looks promising? I'm not sure what cxf does if maxThreads has been
> reached: https://cxf.apache.org/docs/jetty-configuration.html
> And then I find this:
> https://stackoverflow.com/questions/62826590/how-does-jetty-handle-threads-and-threadpools-it-uses-too-much-memory
> and think it would be more straightforward to implement at the Tika layer.
> LOL...
>

Indeed, sounds like controlling the number of threads won’t automatically
trigger a busy reply to clients anyway.

Reply via email to