Right, warmup requests should only cover idempotent endpoints and typically
only reads.

On Fri, Sep 16, 2016 at 3:05 AM, Andrejs Jermakovics <
andrejs.jermakov...@gmail.com> wrote:

> Thank you for you suggestions!
> Warmup requests make a lot of sense and are quite easy to do but we have
> to be careful not to produce any side effects.
>
> On Wednesday, September 14, 2016 at 6:03:57 PM UTC+1, Evan Meagher wrote:
>>
>> More generally, it's not uncommon for server programs on the JVM to
>> benefit from some initial "warmup" workload to allow for classloading, JIT
>> compilation, and other runtime optimizations to kick in before serving
>> external traffic. A good way to go about doing this is to send a series of 
>> warmup
>> requests <https://cloud.google.com/appengine/docs/java/warmup-requests/>
>> through the application before opening the server port to external traffic.
>>
>> One way to achieve this might be to add a hook to AbstractServerFactory
>> to allow an application to provide a `Consumer<Handler>` that the
>> application context handler could be applied to. The provided function
>> could then issue warmup requests directly to the application servlet
>> handler.
>>
>> I may be missing some caveat about the pre-start portion of Jetty's
>> server lifecycle. More research would be necessary if someone chose to
>> pursue this feature.
>>
>> On Wed, Sep 14, 2016 at 9:21 AM, Tatu Saloranta <tsalo...@gmail.com>
>> wrote:
>>
>>> One aspect that would be easy enough to support is warming up of
>>> data-binding library, Jackson. Ideally it'd be done by exercising
>>> ObjectMapper that DropWizard uses (and preferably with actual types used
>>> for serialization/deserialization); but even just creating an instance
>>> and doing simple ser/deser can help with JVM aspects.
>>>
>>> I don't know how big part of first-call overhead is from data-binding
>>> but I know it is non-trivial relative to steady-state overhead (first call
>>> taking hundreds of milliseconds, potentially, steady state fraction of a
>>> millisecond)
>>>
>>> -+ Tatu +-
>>>
>>>
>>> On Wed, Sep 14, 2016 at 2:06 AM, Andrejs Jermakovics <
>>> andrejs.j...@gmail.com> wrote:
>>>
>>>> Hi all,
>>>>
>>>>
>>>> Is there any way to 'warm-up' dropwizard before handling requests?
>>>>
>>>> What we're observing is that the first request to an endpoint is quite
>>>> slow but all the subsequent ones are fast. I assume that some
>>>> initialization is happening the first time you hit an endpoint but we
>>>> couldn't find a way to trigger it.
>>>> Any ideas would be much appreciated!
>>>>
>>>>
>>>> Thank you,
>>>> Andrejs
>>>>
>>>> --
>>>> You received this message because you are subscribed to the Google
>>>> Groups "dropwizard-user" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send
>>>> an email to dropwizard-us...@googlegroups.com.
>>>> For more options, visit https://groups.google.com/d/optout.
>>>>
>>>
>>> --
>>> You received this message because you are subscribed to the Google
>>> Groups "dropwizard-user" group.
>>> To unsubscribe from this group and stop receiving emails from it, send
>>> an email to dropwizard-us...@googlegroups.com.
>>> For more options, visit https://groups.google.com/d/optout.
>>>
>>
>>
>>
>> --
>> Evan Meagher
>>
> --
> You received this message because you are subscribed to the Google Groups
> "dropwizard-user" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to dropwizard-user+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>



-- 
Evan Meagher

-- 
You received this message because you are subscribed to the Google Groups 
"dropwizard-user" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to dropwizard-user+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to