Master site with +20 satellites

Hello Forum !

Our CMK Master is very slow when we Active Changes
Most probably related to many Distributed sites (satellites as we prefer to cal them)

Can we speedup by simple enabling persistent connections on all sites ?
AFAIK this will keep a connection open to every sites and must have some impact on the master memory usages, but if we can gain 40% speed it will improve the workspeed of the engineers who need to activate changes a lot while working on migration projects

Or any other tip is welcome :slight_smile:

Hi,

please take a look at these two topics:

Furthermore you can analyze the configuration by going to Setup → Analyze configuration and identify some bottlenecks. To modify the respective settings, I would suggest doing so from Setup → Distributed monitoring and modify the settings accordingly for each site and for each performance metric.

Best

Analyze the config is in this case pretty worthless :slight_smile: it just reports perst conn are enable or not, and to not use when you have proxy enabled but we know that

I want to know if the persitent connection would improve connectivity to all satellites and if the impact on the master site is mem related, anybody else with large setups, i.g. many sites ?

FTR: our master site has nearly no hosts or services in monitorring, all work is done on the distributed sites

Normally only affected sites should be activated and not all. One other useful information would be the CMK version. With 2.4 i significantly lower activation times in my distributed environments.

The persistent connection is more for the livestatus queries is think and not so important for the activate changes.

Tnx Andreas, makes sense it’s lq related

but when you create a local user it’s pushed on the whole env …

the activate changes is pretty slow since it depends on some factors, like the pending changes, activation in progress, network latency between satellite and master etc…
Activate changes over WATO takes also significant time, so is the request over the API.
It is better than older versions, but it is still slow.

depends on version, are you running something like 2.0 its slow.
remember even if you have all your hosts monitored on your pollers (like we call them) the WATO config is on the master site, if you have 100k hosts with rules every activation will push those to all pollers and the poller will decide what to do.

But this has all changed and only the relevant parts are pushed to the relevant servers in distributed monitoring

also activation is done over HTTP(s) so if you have WAN links with proxies, or bw constrains it can be slow.

Hi

can you give me more details pls?

  1. What means slow in seconds?
  2. Which Version are you using?

Cheers
D

takes 2min (I read a few posts where it is acceptible en within limits) the idea was to know if persistent connection could improve stuff
we are still on 2.2