Sensu capacity & Hardware requirements

Hi all,

I have couple questions on Sensu capacity and Hardware requirements.

I’m trying to figure out if there is a known the max capacity for a particular Sensu server instance. In other words how many hots (checks eventually) can a Sensu server hold ?

Also I want to have a quick overview on the Sensu server hardware requirements recommended like filling the

No.** **of hosts

No. checks

No. CPUs

No. CPU Cores

CPU speed

Memory

Hard disk capacity

Network

< 100

** <500

···

**
** ?**

?

** ?**
?
** ?**
** ?**

100 - 200

500 -1000

** ?**
** ?**
?
?
?
** ?**

> 200

>1000

?
** ?**
** ?**
?
?
** ?
**

> 500

>2500

** ?**
** ?**
** ?**
?
?
?


Thanks for your help.

There are many more variables to account for here. How frequently are the checks in the No. checks column running? (once an hour? once a second?)
Maybe a closer metric would be how many “events per second” could a server handle.

Then again, the sensu-server component scales very well with “green” events because it mostly doesn’t do anything, it only needs to execute things when the check is in a failing state.

Then the capacity is a function of how cpu intensive your handlers are, as traditional (non-extension) handlers are executed in a new process.

So then you might ask “how many failing events per second can a single sensu-server instance handle?” Again, it depends on the handlers and filtering applied to the check (occurrences/refresh/etc)

In the end the answer will be so specific to your particular configuration that I couldn’t really guess for you.

Here is the only other related post I could find:

https://groups.google.com/forum/#!topic/sensu-users/Mm63oMzNUs4

···

On Tue, Mar 10, 2015 at 7:16 AM, Taher Ben Salem bensalemtaher@gmail.com wrote:

Hi all,

I have couple questions on Sensu capacity and Hardware requirements.

I’m trying to figure out if there is a known the max capacity for a particular Sensu server instance. In other words how many hots (checks eventually) can a Sensu server hold ?

Also I want to have a quick overview on the Sensu server hardware requirements recommended like filling the

No.** **of hosts

No. checks

**No. CPUs **

No. CPU Cores

CPU speed

Memory

Hard disk capacity

Network

< 100

** <500
**
** ?**
?

** ?**
?
** ?**
** ?**
100 - 200

500 -1000

** ?**
** ?**
?
?
?
** ?**
> 200

>1000

?
** ?**
** ?**
?
?
** ?
**
> 500

>2500

** ?**
** ?**
** ?**
?
?
?


Thanks for your help.

As Kyle pointed out, it will vary greatly by install. He did a great job describing all the nuances.

For the sake of conversation though, we can (have) manage(d) more hosts and checks from a single commodity server than the last line of your grid when everything is working as expected (this assumes an average number of events along with green). You should run at minimum two servers though for process offload. We have a decommission handler that inspects keepalive events to determine a real event or normal daily decommission activities. If we had a single server and had a keepalive storm, the number of events to handle would surely spike the load on the server very high.

I’d stick to commodity sized servers and go horizontal when you need to. Most commodity sized servers are probably even overkill these days. What is the base HP G8 these days, 10 core, 16 GB? IMO, you don’t need a ton of disk; it most you have logs, many smallish check scripts, etc.

-Bryan

···

On Tue, Mar 10, 2015 at 11:24 AM, Kyle Anderson kyle@xkyle.com wrote:

There are many more variables to account for here. How frequently are the checks in the No. checks column running? (once an hour? once a second?)
Maybe a closer metric would be how many “events per second” could a server handle.

Then again, the sensu-server component scales very well with “green” events because it mostly doesn’t do anything, it only needs to execute things when the check is in a failing state.

Then the capacity is a function of how cpu intensive your handlers are, as traditional (non-extension) handlers are executed in a new process.

So then you might ask “how many failing events per second can a single sensu-server instance handle?” Again, it depends on the handlers and filtering applied to the check (occurrences/refresh/etc)

In the end the answer will be so specific to your particular configuration that I couldn’t really guess for you.

Here is the only other related post I could find:

https://groups.google.com/forum/#!topic/sensu-users/Mm63oMzNUs4

On Tue, Mar 10, 2015 at 7:16 AM, Taher Ben Salem bensalemtaher@gmail.com wrote:

Hi all,

I have couple questions on Sensu capacity and Hardware requirements.

I’m trying to figure out if there is a known the max capacity for a particular Sensu server instance. In other words how many hots (checks eventually) can a Sensu server hold ?

Also I want to have a quick overview on the Sensu server hardware requirements recommended like filling the

No.** **of hosts

No. checks

**No. CPUs **

No. CPU Cores

CPU speed

Memory

Hard disk capacity

Network

< 100

** <500
**
** ?**
?

** ?**
?
** ?**
** ?**
100 - 200

500 -1000

** ?**
** ?**
?
?
?
** ?**
> 200

>1000

?
** ?**
** ?**
?
?
** ?
**
> 500

>2500

** ?**
** ?**
** ?**
?
?
?


Thanks for your help.