BGP metrics
Big picture
Use Prometheus configured for Calico Cloud calico-node to monitor the health of BGP peers within your cluster.
Value
Using the open-source Prometheus monitoring and alerting toolkit, you can view time-series metrics from Calico Cloud components in the Prometheus or Grafana interfaces.
Calico Cloud adds the ability to monitor high-level operations between BGP peers in your cluster. By defining a set of simple rules and thresholds, you can monitor peer-to-peer connection health between your nodes as well as the number of routes being exchanged and receive alerts when it exceeds configured thresholds.
Concepts
+-------------------+
| Host |
| +-------------------+ +------------+ +------------+
| | Host |------------->--| | | |--->--
| | +-------------------+ policy | Prometheus | | Prometheus | alert
+-| | Host |----------->--| Server |-->--| Alert- |--->--
| | +-------------+ | metrics | | | manager | mechanisms
+-| | BGP Metrics |-------------->--| | | |--->--
| | Server | | | | | |
| +-------------+ | +------------+ +------------+
+-------------------+ ^ ^
| |
Collect and store metrics. Web UI for accessing alert
WebUI for accessing and states.
querying metrics. Configure fan out
Configure alerting rules. notifications to different
alert receivers.
BGP metric reporting is accomplished using three key pieces:
- BGP Metrics Server
- Prometheus Server
- Prometheus Alertmanager
About Prometheus
The Prometheus scrapes various instrumented jobs (endpoints) to collect time series data for a given set of metrics. Time series data can then be queried and rules can be setup to monitor specific thresholds to trigger alerts. The data can also be visualized (such as using Grafana).
Prometheus Server deployed as part of the Calico Cloud scrapes every configured calico-node target. Alerting rules querying BGP metrics can be configured in Prometheus and when triggered, fire alerts to the Prometheus Alertmanager.
Prometheus Alertmanager (or simply Alertmanager), deployed as part of the Calico Cloud, receives alerts from Prometheus and forwards alerts to various alerting mechanisms such as Pager Duty, or OpsGenie.
About Calico Cloud calico-node
calico-node bundles together the components required for networking containers with Calico Cloud. The key components are:
- Felix
- BIRD
- confd
Its critical function means that it runs on every machine that provides endpoints. A binary running inside calico-node monitors the BIRD daemon for peering and routing activity and reports these statics to Prometheus.
How to
BGP metrics are generated within calico-node every 5 seconds using statistics pulled from the BIRD daemon.
The metrics generated are:
bgp_peers- Total number of peers with a specific BGP connection status.bgp_routes_imported- Current number of routes successfully imported into the routing table.bgp_route_updates_received- Total number of route updates received over time (since startup).
Calico Cloud will run BGP metrics for Prometheus by default. Metrics are directly available on each compute node at http://<node-IP>:9900/metrics.
Refer to Configuring Prometheus for information on how to create a new Alerting rule or updating the scraping interval for how often Prometheus collects the metrics.
BGP peers metric
The metric bgp_peers has the relevant labels instance, status and ip_version. Using this metric, you can identify how many peers have a specific BGP connection status with a given node instance and IP version. This metric will be available as a combination of {instance, status, ip_version}.
Example queries:
- Total number of peers currently with a BGP connection to the node instance “calico-node-1”, with status “Established”, for IP version “IPv4”.
bgp_peers{instance="calico-node-1", status="Established", ip_version="IPv4"}
- Total number of peers currently with a BGP connection to the node instance “calico-node-1”, with status “Down”, for IP version “IPv6”.
bgp_peers{instance="calico-node-1", status="Down", ip_version="IPv6"}
- Total number of peers currently with a BGP connection to any node instance, with a status that is not “Established”, for IP version “IPv4”.
bgp_peers{status!="Established", ip_version="IPv4"}
Valid BGP connection statuses are: "Idle", "Connect", "Active", "OpenSent", "OpenConfirm", "Established", "Close", "Down" and "Passive".
BGP routes imported metric
The metric bgp_routes_imported has the relevant labels instance and ip_version. Using this metric, you can identify how many routes are being successfully imported into a given node instance's routing table at a specific point in time. This number can increase or decrease depending on how BGP rules process incoming routes. This metric will be available as a combination of {instance, ip_version}.
Example queries:
- Computes the per-second rate for the number of routes imported by a specific node instance “calico-node-1” looking up to 120 seconds back (using the two most recent data points).
irate(bgp_routes_imported{instance="calico-node-1",ip_version="IPv4"}[120s])
- Computes the per-second rate for the number of routes imported across all node instances looking up to 120 seconds back (using the two most recent data points).
irate(bgp_routes_imported{ip_version="IPv4"}[120s])
BGP route updates received metric
The metric bgp_route_updates_received has the relevant labels instance and ip_version. Using this metric, you can identify the total number of BGP routes received by a given node over time. This number includes all routes that have been accepted & imported into the routing table, as well as any routes that were rejected as invalid, rejected by filters or rejected as already in the route table. This total number should only increase over time. This metric will be available as a combination of {instance, ip_version}.
Example queries:
- Computes the per-second rate for the number of routes received by a specific node instance “calico-node-1 ” looking up to 5 minutes back (using the two most recent data points).
irate(bgp_route_updates_received{instance="calico-node-1",ip_version="IPv4"}[5m])
- Computes the per-second rate for the number of routes received across all node instances looking up to 5 minutes back (using the two most recent data points).
irate(bgp_route_updates_received{ip_version="IPv4"}[5m])