Even if that plugin can be optimized for performance, it still feels like
an inefficient approach. Users will add and modify devices in Netbox, for
example marking them off-line to remove from monitoring. These changes are
infrequent, and I want Prometheus to respond as fast as possible. So I’m
Thanks, it sounds promising on the Prometheus side.
I’ve actually found a performance issue with the Prometheus plug-in for
Netbox that I was using to provide the HTTP discovery. It hammers the
Netbox database with excessive queries and takes over 30 seconds to respond
with just 2000 targets. So,
Thank you for clarification. I was interested in If there are any
disadvantages If the amount of CPU cores is too high maybe because of
Overhead to share the load.
Good to know i can scale it easily If i run it on VMs
Ben Kochie schrieb am Samstag, 13. Januar 2024 um 10:51:49 UTC+1:
> No, Go
No, Go is not specifically limited to a number of cores. For the exporters,
they should scale vertically just fine as well as horizontally.
The only limit I've seen is how well the SNMP exporter's UDP packet
handling works. IIRC you may run into UDP packets per second limits before
you run into
Just to clarify: I picked "4 cores" out of thin air just as an example to
work through, same as I picked 15 second scrape interval and 150ms per
scrape.
On Saturday 13 January 2024 at 09:34:21 UTC Brian Candler wrote:
> One reason is you may already have eight 4-core servers lying around.
>
>
One reason is you may already have eight 4-core servers lying around.
If it's a VM then of course you can just scale up to the largest instance
size available, before you need to go to multiple instnaces.
On Saturday 13 January 2024 at 00:20:10 UTC Alexander Wilke wrote:
> Hello,
> sorry to
Hello,
sorry to hijack this thread a little bit but Brian talks about "4 CPU
cores" and Ben says "scale horizontally".
Just for interest - why not just use 8, 16, or 32 CPU cores? Is Go limited
at a specific CPU amount or is there a disadvantage to have to many cores?
I think if someone is
Those sound like reasonable amounts for those exporters.
I've heard of people hitting thousands of SNMP devices from the
snmp_exporter.
Since the exporters are in Go, they scale well. But if it's not enough, the
advantage of their design means they can be deployed horizontally. You
could run
I'm curious if anyone has experimented to find out how many targets can
reasonably be scraped by a single instance of blackbox and snmp exporters.
I know Prometheus itself can handle tens of thousands of targets, but I'm
wondering at what point it becomes necessary to split up the scraping.
I'll
9 matches
Mail list logo
