Hi Alexander,
Since Observium uses MySQL as its back end database and RRD for metrics, redundancy options are based on what you can do with MySQL and filesystem replication. Observium doesn't actually need to be involved or even aware of that at all. The only consideration is that filesystem replication would interact poorly with rrdcached, if rrdcached is used for remote pollers.
Master/master replication of MySQL and the filesystem containing the RRDs would permit separate GUI instances at each location without any latency impact.
I can't really speak to the scalability of this solution as I've not seen it used, but I don't think there are any major hurdles, assuming the replication is reliable.
We'd generally be willing to help try to improve performance/scalability on large Enterprise installs.
There's no ability to "merge" two separate instances into a single UI. I'm not even sure I'm brave enough to imagine how complex that would be!
Thanks,
Adam.
From: observium observium-bounces@observium.org On Behalf Of Joelly, Alexander via observium Sent: 24 January 2019 00:22 To: observium@observium.org Cc: Joelly, Alexander Alexander.Joelly@magna.com Subject: [Observium] Observium Enterprise
Hello,
Our company is split into two major regions (EU & NA) where we have a central datacenter in each region, all sites in the particular region is connected to it's regional datacenter and the two datacenters are red. connected together. As I understand with the enterprise version multiple pollers can be deployed within the topology to scale vertically where in our scenario we would place them in our datacenters as everything is reachable from there (DC and WAN equippement on remote sites).
How would we monitor both regions? Our concern is that the latency between the datacenters (~120ms) could cause some issues and might slow in response if the gui,rrd and mysql servers are hosted in one region.
Can the rrd and mysql servers deployed redundant - one node in EU and the second in NA and also each region a local gui instance? Or is it possible to run a separate instance in each region and have a common gui with visibility of all devices? We possibly might also monitor the LAN equipment of the sites in the future - that might add a couple of thousands devices per region (no real numbers known yet)
We have a network team and service desk in each region, where we have about 1.500 datacenter and WAN devices in each region to monitor in the first step - we are calculating to have around 30 people from each region using the gui daily.
Does someone have such a setup?
We setup the CE for testing and we would go ahead using it globally, but not sure how the setup should/would like for that and we haven't found anything in that direction.
Thanks, Alex