Hi,
Thanks for the update, fixes and sorry for late response. We've been busy with
a release for which we had to overcome this performance issue by migrating the
expressions and using "ref:" language.
We'll run some trials of the same code in the near future of the 3.4.5-SNAPSHOT
version if I understood correctly and will post an update.
Definitely we won't migrate to 3.6.x, we'll stick to 3.4.x for some time.
Thank you, Corneliu
-----Original Message-----
From: Claus Ibsen <claus.ib...@gmail.com>
Sent: Monday, October 5, 2020 15:43
To: users@camel.apache.org
Subject: Re: Performance regression with bean and ognl expressions in Simple
language version 3.4.x
Hi Corneliu
Thanks for the reproducer. We have fixed a bunch of stuff and you are welcome
to try with SNAPSHOT and do profiling again and report back if you can find
improvements too.
Or if you dont have time then test again when Camel 3.6.0 comes out later in
October
On Thu, Sep 17, 2020 at 2:32 AM Corneliu Chitic
<corneliu.chi...@computaris.com.invalid> wrote:
>
> Hi,
>
> we've identified a performance regression while running same code with Apache
> Camel 3.4.3 + Spring Boot vs Apache Camel 2.24.2 with Spring framework 5.1.9.
> We've migrated one application to this LTS version and we face this impact.
> The main bottleneck is the synchronized block from:
> org.apache.camel.impl.engine.AbstractCamelContext.resolveLanguage(String).
> The root cause is the time spent to validate Simple expressions when using
> bean language (${bean:name?method=something}) or OGNL like calls to POJO
> methods (${exchangeProperty.pojo.method}). According to the stack traces the
> new version spends time to allocate the bean + full setup of it. Blocking
> times are quite high (average 100ms, max could be ~300ms) and as the number
> of parallel processing threads increases it goes up steadily.
>
> Has anything changed in version 3.x (or more precisely 3.4.x)? The changelogs
> and upgrade tutorial didn't suggested anything in this area.
> Is there any configuration flag that would allow us to switch back to version
> 2.x mode of working for this functionality?
>
> We have run repeated trials and have consistent results with both versions;
> we have a project setup to demo this and also some Java Flight recordings for
> comparison. I don't think I can attach anything to this maillist, please let
> me know how I can provide any additional input if needed.
>
> Thank you, Corneliu
> This email is subject to Computaris email terms of use:
> https://www.computaris.com/email-terms-use/
--
Claus Ibsen
-----------------
http://davsclaus.com @davsclaus
Camel in Action 2: https://www.manning.com/ibsen2
