Any feedback on this? On Tue, Nov 27, 2018 at 4:14 PM Ivan Junckes Filho <ivanjunc...@gmail.com> wrote:
> Hey guys I did some improvements on this example: > > - Fixing payload in test after the fix of a jsonb regression > - Replace Webclient with WebTarget > - Fix to use top level properties with versions > - Better assertion by key type > > Could you please review? > https://github.com/apache/tomee/pull/221 > > On Tue, Nov 20, 2018 at 8:35 PM Romain Manni-Bucau <rmannibu...@gmail.com> > wrote: > >> Well it was not a big issue bit something to take care. The pr - not yet >> merged - is way more impacting users and kind of justifies to not use MP >> today :(. >> >> Le mar. 20 nov. 2018 22:39, César Hernández Mendoza <cesargu...@gmail.com >> > >> a écrit : >> >> > Hi Romain, >> > >> > Did you test passing accept header or just in the browser which sends >> json >> > > and text as acceptable value so resolution can be not that >> deterministic >> > > using it - but this is the request so all fine ;). >> > >> > >> > Yes, yesterday my test was performed using safari and that was causing >> the >> > behaviour I describe in my previous email. >> > Today I build tested again and did some test using curl and the Accept >> > header and I didn't find any issue. Thanks for the help. >> > >> > The last side note is prometheus does not support monotonic=false (small >> > > glitch in the spec). >> > >> > Interesting, It seems this is been fixed for microprofile metrics 2.x >> > release [1] [2] as we are speaking. >> > >> > [1] https://github.com/eclipse/microprofile-metrics/pull/309 >> > [2] https://github.com/eclipse/microprofile-metrics/issues/290 >> > >> > >> > El lun., 19 nov. 2018 a las 23:52, Romain Manni-Bucau (< >> > rmannibu...@gmail.com>) escribió: >> > >> > > Hi César, >> > > >> > > Did you test passing accept header or just in the browser which sends >> > json >> > > and text as acceptable value so resolution can be not that >> deterministic >> > > using it - but this is the request so all fine ;). >> > > >> > > The last side note is prometheus does not support monotonic=false >> (small >> > > glitch in the spec). >> > > >> > > Le mar. 20 nov. 2018 03:12, César Hernández Mendoza < >> > cesargu...@gmail.com> >> > > a écrit : >> > > >> > > > Hi Ivan, nice job with this example. >> > > > >> > > > I found another potential issue with the metrics in TomEE, If you >> > remove >> > > > the `monotonic = true` from the @Counted annotation, then the >> format of >> > > the >> > > > endpoint: http://localhost:8080/rest-mp-metrics/metrics is no >> longer >> > > > Prometheus but JSON instead. >> > > > >> > > > >> > > > Without `monotonic = true` >> > > > >> > > > >> > > > >> > > >> > >> {"application":{"message_counter":0},"vendor":{"startTime":1542679115312},"base":{"classloader.totalLoadedClass.count":8971,"thread.count":39,"classloader.currentLoadedClass.count":8971,"jvm.uptime":103050,"gc.PS >> > > > >> > > > >> > > >> > >> MarkSweep.count":2,"thread.max.count":39,"memory.committedHeap":369098752,"gc.PS >> > > > >> > > > >> > > >> > >> Scavenge.count":7,"cpu.availableProcessors":4,"thread.daemon.count":38,"classloader.totalUnloadedClass.count":8971,"memory.maxHeap":3817865216,"memory.usedHeap":69535960,"gc.PS >> > > > MarkSweep.time":141,"gc.PS Scavenge.time":103}} >> > > > >> > > > >> > > > >> > > > With `monotonic = true` >> > > > >> > > > # TYPE application:message_counter counter >> > > > application:message_counter 0.0 >> > > > # TYPE base:classloader_total_loaded_class_count counter >> > > > base:classloader_total_loaded_class_count 8839.0 >> > > > # TYPE base:thread_count counter >> > > > base:thread_count 39.0 >> > > > # TYPE base:classloader_current_loaded_class_count counter >> > > > base:classloader_current_loaded_class_count 8847.0 >> > > > # TYPE base:jvm_uptime_seconds gauge >> > > > base:jvm_uptime_seconds 0.006427 >> > > > # TYPE base:gc_ps_mark_sweep_count counter >> > > > base:gc_ps_mark_sweep_count 2.0 >> > > > # TYPE base:gc_ps_scavenge_count counter >> > > > base:gc_ps_scavenge_count 7.0 >> > > > # TYPE base:memory_committed_heap_bytes gauge >> > > > base:memory_committed_heap_bytes 3.70147328E8 >> > > > # TYPE base:thread_max_count counter >> > > > base:thread_max_count 39.0 >> > > > # TYPE base:cpu_available_processors gauge >> > > > base:cpu_available_processors 4.0 >> > > > # TYPE base:thread_daemon_count counter >> > > > base:thread_daemon_count 38.0 >> > > > # TYPE base:classloader_total_unloaded_class_count counter >> > > > base:classloader_total_unloaded_class_count 8847.0 >> > > > # TYPE base:memory_max_heap_bytes gauge >> > > > base:memory_max_heap_bytes 3.817865216E9 >> > > > # TYPE base:gc_ps_mark_sweep_time_seconds gauge >> > > > base:gc_ps_mark_sweep_time_seconds 1.19E-4 >> > > > # TYPE base:memory_used_heap_bytes gauge >> > > > base:memory_used_heap_bytes 6.7481504E7 >> > > > # TYPE base:gc_ps_scavenge_time_seconds gauge >> > > > base:gc_ps_scavenge_time_seconds 1.09E-4 >> > > > # TYPE vendor:start_time counter >> > > > vendor:start_time 1.542679428814E12 >> > > > >> > > > >> > > > This is not stated on the MP metrics spec. So I wonder if TomEE >> should >> > > > provide a set of properties to avoid these random behaviors. >> > > > >> > > > >> > > > El lun., 19 nov. 2018 a las 13:51, Ivan Junckes Filho (< >> > > > ivanjunc...@gmail.com>) escribió: >> > > > >> > > > > Hey guys, just to let you know that I created a Microprofile >> Metrics >> > > > > Example under Tomee. >> > > > > >> > > > > I am still working on that and I will provide more test scenarios. >> > > > > >> > > > > Here is the PR if you want to review the work so far: >> > > > > https://github.com/apache/tomee/pull/203 >> > > > > >> > > > > Feel free to merge if you think this can be done incrementally. >> > > > > >> > > > >> > > > >> > > > -- >> > > > Atentamente: >> > > > César Hernández Mendoza. >> > > > >> > > >> > >> > >> > -- >> > Atentamente: >> > César Hernández Mendoza. >> > >> >