Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [jetty-dev] Information regarding Jetty HTTP Client properties

Thanks for the details Simone.

The Java Doc that mentions this is available @ http://download.eclipse.org/jetty/stable-9/apidocs/org/eclipse/jetty/client/HttpClient.html#setMaxConnectionsPerDestination--

It also mentions that this needs to be set to a higher value in case we need to load tests, but our use case is a PROD use case, and not a test use case, so was confused if the high setting value would lead to issues/instability.
We would anyways, try setting a higher value and do some tests to confirm if this scales or not.

Regards
Neha

On Thu, Apr 20, 2017 at 2:56 AM, Simone Bordet <sbordet@xxxxxxxxxxx> wrote:
Hi,

On Wed, Apr 19, 2017 at 9:06 PM, Neha Munjal <neha.munjal3@xxxxxxxxx> wrote:
>> In that case, the purpose of setting this property would help in
>> identifying a request
> that carries a bad host name before even trying to establish a connection
> and result in exceptions like UnknownHostException.
> Can you please clarify this?

It is a DNS timeout. I would not give it more meaning than what it actually is.
Whether it is caused by a bad host name, a bad network, a bad ISP,
etc. is not really relevant.

> Also, we have a use case wherein we have a requirement to send approximately
> 10,000 HTTP/2 requests per second to a target end point.
> We plan to configure the client object with 1 connection and a high value of
> the number of queued requests.

I don't think this is going to fly.
For 10k requests/s you probably need more than 1 connection to the
server, otherwise your requests are going to spend a lot of time
queued.

> (Also, as mentioned in the Java Docs, it is advised to set the number of
> connections property to a maximum of 2.)

Where is this ?

> We do have a throttling mechanism on our end to send approx to trigger
> 10,000 requests from the client end.
> Is there a correlation that we could use to set the number of queued
> requests with 1 connection with the throughput we want to achieve to be able
> to send these requests successfully without running into request rejection
> failures or in such a scenario, can we configure the client object with a
> higher number of connection.
> Please let me know your recommendation.

This is really complicate matter, it depends on a lot of factors,
among which network latency, request/response sizes, etc.

I recommend that you look into our load generator, since typically the
client is the one that is subject to the greater stress and cannot
keep up with the server.
One single client generating 10k requests/s in steady state, I think I
have still to see it; you probably will need more than 1.
https://github.com/jetty-project/jetty-load-generator

If you have this hard requirement of 10k requests/s, I suggest you
contact Webtide for commercial support, as we have done this kind of
things in the past for other customers.
https://webtide.com/

--
Simone Bordet
----
http://cometd.org
http://webtide.com
Developer advice, training, services and support
from the Jetty & CometD experts.
_______________________________________________
jetty-dev mailing list
jetty-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/jetty-dev


Back to the top