Hi Bo, Before reviewing the details, I would like to share a high-level idea.
As I understand, the proposal wants to provide metrics for users to monitor if the subscription is consumed in a real-time manner. Can we just add a metric for (dispatch time - publish time)? So that you can know how long it takes from publishing and consumption. So that you don't need to read the message, which might introduce IO cost if missed from the cache. But it can also indicate if the subscription is consuming data in real-time. But it can't resolve a case that the consumer suddenly stops consuming. All the previous messages are consumed with very low latency. If that could be a problem, we can introduce another metric, e.g., subscription_idle_time. Or maybe we don't need it. We can check if the backlog continues going up or no message delivery throughput of the subscription, but there are backlogs. WDYT? Regards, Penghui On Thu, Aug 10, 2023 at 7:00 PM Qiang Huang <qiang.huang1...@gmail.com> wrote: > +1 > > houxiaoyu <houxia...@apache.org> 于2023年7月24日周一 20:00写道： > > > +1 > > > > Thanks > > Xiaoyu Hou > > > > 丛搏 <congbobo...@gmail.com> 于2023年7月24日周一 19:58写道： > > > > > Hi, Pulsar Community > > > > > > I opened a new PIP design PR. > > > > > > https://github.com/apache/pulsar/pull/20859 > > > > > > Thanks, > > > Bo > > > > > > > > -- > BR, > Qiang Huang >
