杨杰, yangjie
杨杰, yangjie
may is the reason that jsonpb package so slow.
After check the grafana, i find that the queue length is even higher after i change it to batch queue. So maybe i was wrong. I will try to republish...
I made a test use prod environment. My machine is 12 core, 24 gb memory, amd64. I run collector with param --collector.num-workers=12,--collector.queue-size=1000000 and env GOGC=1000. Here is the grafana metric...
Yes, i keeping the same queue capacity both in batch queue and not batch queue. But i modify the behavior of Produce to let the queue.Size() means the real number...
after upgrade the --storage-active-memory from 512mb to 50gb, seems work right
for now, tha parca ui provide a target html, which enable we collector a profile immediately through go http pprof handler.  if the parca agent could...
> 嘿@zdyj3170101136，我不确定我是否理解该请求，听起来这个想法是通过 API 从代理通过 PID触发+**拉取配置文件？**也许还有另一个 API 可以列出节点上的进程？基本上，按需而不是连续地分析选定的流程？ > > 即将发布的 0.10 版代理将允许在系统范围内进行分析，所有正在运行的进程的配置文件（过滤计划：[parca-dev/parca-agent#690](https://github.com/parca-dev/parca-agent/issues/690)）将**被推**送到 Parca：[polarsignals.com/blog/posts/2022/ 08/24/系统范围的分析](https://www.polarsignals.com/blog/posts/2022/08/24/system-wide-profiling/) > > 代理 UI 的更新正在进行中：[parca-dev/parca-agent#767](https://github.com/parca-dev/parca-agent/pull/767)。它恰好添加了一个`/query`端点，但我认为目前没有任何计划在 Parca 中使用这样的端点。端点不会触发配置文件的创建，它会等待分析循环生成的下一个配置文件并在将其发送到 Parca 之前将其捕获。 > > 不过，您可能对以下 2 个功能请求感兴趣： > >...
we had support immediately profiling base on parca. here is a page which use can input ip, port and endpoint.  and then user could click on...
still not merged?
when i read the code. i find the calculation of sample event and sample space is different.   with small event buffer, we...