@tim for a first analysis is fine to use JVisualVM, but I recommend
async-profiler (or perf tools) because it shows GC activity, Kernel
activity (on the networking layer and disk) and, if they are using a native
SSL provider, the native methods cost of it (that often is huge). In
addition, JVisualV
JVisualVM, which ships with the Oracle JDK (not sure about OpenJDK) is
another option. Be sure to use the CPU sampler, not the CPU profiler, if
you can't afford to add a significant load to the process (e.g. if this is
an operational host).
Tim
On Wed, Apr 22, 2020, 4:21 PM Francesco Nigro wrote
I mean not just GC logs but CPU profiling (async-profiler is a mixed
native-java profiler). Without something like that I cannot see how to help.
Il mer 22 apr 2020, 23:43 Ybello ha scritto:
> Hi thanks for the answers
> Yes we did almost every memory analysis tools and we didn't find any
> symp
Hi thanks for the answers
Yes we did almost every memory analysis tools and we didn't find any
symptoms
It defentley something with the Artemis Broker...
We change the network Docker host implementation as well , insant of using
--proxy-docker we change it to --host network and still not luck
8bho
Have you tried profiling with async-profiler (or perf tools) where the CPU
time is spent? It could be either lack of memory on the heap (hence the GC
is doing much more work to free up memory continuously) or whatever
possible reason...
Il mer 22 apr 2020, 23:04 Ybello ha scritto:
> Hello everyo
Hello everyone
Recently we suffering from bad performance that cause our Artemis to be on
100% of cpu , load average is 12.8
Our Artemis is sitting on o e gcp vm based on Docker image, the version that
we running is 2.10
I really want need your help to understanding the problems, we've been
trying