26 lines
2.1 KiB
Plaintext
26 lines
2.1 KiB
Plaintext
[role="xpack"]
|
|
[[uptime-deployment-arch]]
|
|
== Deployment Architecture
|
|
|
|
There are multiple ways to deploy Uptime and Heartbeat.
|
|
Use the information in this section to determine the best deployment for you.
|
|
A guiding principle is that an outage that takes down the service being monitored should not also take down Heartbeat.
|
|
You want Heartbeat to be functioning even when your service is not, so the guidelines here help you maximise this possibility.
|
|
|
|
Heartbeat is generally run as a centralized service within a data center.
|
|
While it is possible to run it as a separate "sidecar" process paired with each process/container, we recommend against it.
|
|
Running Heartbeat centrally ensures you will still be able to see monitoring data in the event of an overloaded, disconnected, or otherwise malfunctioning server.
|
|
|
|
For further redundancy, you may want to deploy multiple Heartbeats across geographic and/or network boundaries to provide more data.
|
|
Specify Heartbeat's observer {heartbeat-ref}/configuration-observer-options.html[geo options] to do so. Some examples might be:
|
|
|
|
* **A site served from a content delivery network (CDN) with points of presence (POPs) around the globe:**
|
|
In this case you may want to have multiple Heartbeat instances at different data centers around the world checking to see if your site is reachable via local CDN POPs.
|
|
* **A service within a single data center that is accessed across multiple VPNs:**
|
|
Set up one Heartbeat instance within the VPN the service operates from, and another within an additional VPN that users access the service from.
|
|
Having both instances will help pinpoint network errors in the event of an outage.
|
|
* **A single service running primarily in a US east coast data center, with a hot failover located in a US west coast data center:**
|
|
In each data center, run a Heartbeat instance that checks both the local copy of the service and its counterpart across the country.
|
|
Set up two monitors in each region, one for the local service and one for the remote service.
|
|
In the event of a data center failure it will be immediately obvious if the service had a connectivity issue to the outside world or if the failure was only internal.
|