Hi Matija,
Thank you so much for the reply. The zipking seems to be a very useful
tool and I'll take a close look at it.
Meanwhile, I know how to use the Jconsole to get the information exposing
to JMX. However, for the Jconsole, I need to click the refresh button to get
the updated value for each metrics. I'm wondering whether there's a convenient
way that I can print the time-series value out through the terminal of some
metrics.
Thanks!
Jun
From: [email protected]
Date: Tue, 30 Aug 2016 21:08:45 +0200
Subject: Re: How to get information of each read/write request?
To: [email protected]
Hi Jun,
If you are looking to track each request zipking is your best bet. The last
pickle has a blog about tracing using zipkin.Regarding the stats you see in
nodetool did you check the metrics package in cassandra and what it exposes
over JMX?
Regards,Matija
On Tue, Aug 30, 2016 at 7:20 PM, Jun Wu <[email protected]> wrote:
Hi there,
I'm very interested in the read/write path of Cassandra. Specifically, I'd
like to know the whole process when a read/write request comes in.
I noticed that for reach request it could go through multiple stages. For
example, for read request, it could be in ReadStage, RequestResponseStage,
ReadRepairStage. For each stage, actually it's a queue and thread pool to serve
the request.
First question is how to track each request in which stage. Also I'm very
interested int the waiting time for each request to be in the queue, also the
total queue in each stage. I noticed that in nodetool tpstats will have this
information. However, I may want to get the real-time information of this, like
print it out in the terminal.
I'm wondering whether someone has hints on this.
Thanks in advance!
Jun