Collecting and presenting host health status
Hi
Usually we measure the VPS status using grafana or other system to present cpu, ram, disk, etc, but is there a way to collect information/signs (and what information) about the hosts where our VPS runs so we can measure the host health?
We usually get a good VPS with some nice specs, but the performance is not up to the expectations or the host is getting full and the performance is getting worse, so this information would allow us to take some actions to correct the problem.
Thanks
Comments
Elasticsearch+logstash on host for centralized monitoring, then heartbeat/filebeat lightweight client on VPS
https://www.elastic.co/guide/en/beats/heartbeat/current/index.html
----
oh fugg I am not entirely clear whether you have access to the Host or not
Fuck this 24/7 internet spew of trivia and celebrity bullshit.
No I don't have, as a client I would like to gather realtime information that I could use to evaluate the host health.
I know that CPU steal time and disk read/write (*) are signs I can measure inside a VPS that gives me some information about how dense the host is.
Are there more signs I can use that would let me track the host health?
(*) after understanding how my apps perform and use the disk like, intensive tasks or database read/write.
nope, I'm out of ideas. without host access any information is very limited.
outside of heartbeats or elastic stuff, probably worth checking is hetrixtools agent. they're using it to gather data, you might able to salvage one or two functions if it's interesting
Fuck this 24/7 internet spew of trivia and celebrity bullshit.
If you're running something that sensitive then you shouldn't be using VPS from LE* space imo. Either go for a dedi or a bigger cloud provider since they tend to be a bit more consistent (for better or worse)
What type of virtualization is used? As we know in the case of KVM, if there is no access to the host machine, then it will not be able to receive data.
I think you might read a little closer. He explains he's trying to infer information based on what he can see inside the VM. High cpu steal tells you the CPU is oversold. Low IOPs tells you the disk is either spinning rust or oversold. He's looking for other ideas.