In our check_mk EE, there's around 550 hosts and 14500 services being
monitored (mix of check_mk-native checks, classical active Nagios checks,
and SNMP). The server (which is handling quite a bit of syslog-receiving
as a side activity) has a load between 1 and 3. This server has ten
physical cores in one socket (20 hyperthreads), and its CPU utilization
is on average 12%.
Not sure if you’re running Check_MK within OMD, if not that should help considerably with it’s out of the box optimizations.
I run around 990 hosts, 25000 services for about 350 checks/s on an ESXi 5.5 host with only 12Gb mem and 4 CPU cores. CPU Utilization was at around 30% with load averages between 1 - 5. I’ve recently upgraded from OMD+Check-mk raw 1.2.6 to EE+CMC. This was a considerable performance improvement with load averages around .5 with usage at 15%. Disk writes of only around 2MB/s avg is fairly reasonable considering the amount of performance data.
-Cory
On Mon, Oct 5, 2015 at 12:24 PM, Gerardo Ferreyra raptorg83@gmail.com wrote:
Hi guys.
I have check_mk running over a RHEL 6.6 (minimal installation). 170 hosts.
Some with snmp, 90% with check_mk agent, 30 Esx hosts, others (windows NT) with another types of monitoring.
170 hosts!!!
This is a VM over VMware 5.5 (physical: Dell R815, Opteron 6300).
It’s using 100% of theis cpu’s, 2 cores.
I put it 2 more cores. Now, 50% use (of course :))
I thought this was a lightweight monitoring tool. CPU won’t bother, etc.
“Way better than other types of monitoring”
I need to monitor 1000 more hosts. This is going up? I will need 10 more cores?
This is something that concerns me, very much.
Really, i need some data.
Does check_mk run extremeley better on… for example:
what Troels was writing is that he uses the Enterprise Edition of Check_MK with the new build CMC (MicroCore). This core is way more powerful to handle many checks.
The one thing what is also important is your used hypervisor and storage system.
With classic Nagios core i have some systems with nearly the same setup as Troels but with way higher CPU load and utilization. That is the biggest difference between Nagios and CMC.
Not sure if you’re running Check_MK within OMD, if not that should help considerably with it’s out of the box optimizations.
I run around 990 hosts, 25000 services for about 350 checks/s on an ESXi 5.5 host with only 12Gb mem and 4 CPU cores. CPU Utilization was at around 30% with load averages between 1 - 5. I’ve recently upgraded from OMD+Check-mk raw 1.2.6 to EE+CMC. This was a considerable performance improvement with load averages around .5 with usage at 15%. Disk writes of only around 2MB/s avg is fairly reasonable considering the amount of performance data.
-Cory
On Mon, Oct 5, 2015 at 12:24 PM, Gerardo Ferreyra raptorg83@gmail.com wrote:
Hi guys.
I have check_mk running over a RHEL 6.6 (minimal installation). 170 hosts.
Some with snmp, 90% with check_mk agent, 30 Esx hosts, others (windows NT) with another types of monitoring.
170 hosts!!!
This is a VM over VMware 5.5 (physical: Dell R815, Opteron 6300).
It’s using 100% of theis cpu’s, 2 cores.
I put it 2 more cores. Now, 50% use (of course :))
I thought this was a lightweight monitoring tool. CPU won’t bother, etc.
“Way better than other types of monitoring”
I need to monitor 1000 more hosts. This is going up? I will need 10 more cores?
This is something that concerns me, very much.
Really, i need some data.
Does check_mk run extremeley better on… for example:
