upbeat summit
hey guys,
So recently, i've been reading on an article from a consultant i talk with regarding Cassandra performance.
he has been writing an article about hardware profiling with bcc-tools https://rustyrazorblade.com/post/2023/2023-11-14-bcc-tools/
he shows awesome tools like cachestat, biolatency, xfsslower, and xfsdist
Are you guys using this kind of information and showing them in the dashboard?
muted violet
Hello
No, we do not use these tools directly, because they need dependencies we did not want to have.
but, the general idea of these monitoring tools and ours is exactly the same.
We can resume what we are doing like this:
We also deliver for users options that can be used to modify collection, but they cannot change directly and recompile like BCC.
You are not missing metrics when compare boths, because we are trying to deliver exactly the same
Any doubt, please, let us know.
upbeat summit
totally get you guys don't want to use dependencies.
in this case, Jon that i link to, reference to the Bcc-tools and he straight up writes:
For a little background, bcc-tools is a collection of utilities that expose underlying kernel statistics through eBPF. eBPF is an amazing technology that lets us tap into what’s going on in the Linux kernel with negligible performance overhead. Fun stuff!
So, if those tools, can get the info out of the kernel through eBPF, then it should be possible for netdata too.
and maybe those cli tools i list (through him) is already shown in netdata, i just don't know what mapping they are then.
so maybe i'm also just fishing for the "mapping" info, when looking at his blog post it shows information, but i have no idea what that could be within netdata.
muted violet
I am working now in a set of changes for eBPF to be more clear the metrics we are collecting.
Both, Iovisor and US collect the same metrics, but sometimes names are different. I am renaming our metrics to let more clear
I remember that our cachestat, dc, fd, VFS were collecting exactly the same metrics
I have to confirm, because like them we never stop, so probably they added or remove something and we are not in the same page only because of this.
Our goal is to give you the exactly the same metrics, but without necessity to install packages to compile eBPF binaries.
I remember that years ago users had to install kernel-headers and clang to compile and use. This was the reason we opt to make 2 differnt repos to do the hard work and delivery codes ready to use.
upbeat summit
ya