A Prometheus exporter which uses XenServer's XAPI to derive various metrics.
xapi_exporter [-config <configFile>]
If no configFile is specified, xapi_exporter will attempt to read its configuration from "xapi_exporter.yml" in the working directory.
YML config file should look something like:
bindaddress: ":9290"
namespace: "xenstats"
timeoutlogin: 15
pools:
"xen-pool-a":
- "xen-pool-a-host-1"
- "xen-pool-a-host-2"
"xen-pool-b":
- "xen-pool-b-host-1"
- "xen-pool-b-host-2"
- "xen-pool-b-host-3"
- "xen-pool-b-host-4"
auth:
username: "your_xapi_user"
password: "your_xapi_password"
Note that you can optionally select which metrics are reported by including an "enabledmetrics" key to an array of strings. Without the "enabledmetrics" option, all metrics are enabled. If one or more metrics are specified with this option, only those metrics will be reported.
enabledmetrics:
- "memory_free"
- "memory_total"
The above would only generate memory_free and memory_total metrics.
Below are available metrics.
These metrics are relevant to XenServer pools, and will have a "pool" label to distinguish them.
ha_allow_overcommit: If set to false then operations which would cause the Pool to become overcommitted will be blocked.
ha_enabled: true if HA is enabled on the pool, false otherwise
ha_host_failures_to_tolerate: Number of host failures to tolerate before the Pool is declared to be overcommitted
ha_overcommitted: True if the Pool is considered to be overcommitted i.e. if there exist insufficient physicalk resources to tolerate the configured number of host failures
wlb_enabled: true if workload balancing is enabled on the pool
These metrics are relevant to XenServer hosts, and will have a "host" label to distinguish them.
cpu_pct_allocated: percent vCPUs over total CPUs on this host
cpu_count: the number of physical CPUs on the host
memory_pct_allocated: percent used memory over total memory on this host
memory_free: Free host memory (bytes)
memory_total: Total host memory (bytes)
resident_vcpu_count: count of vCPUs on VMs running on this host
resident_vm_count: count of VMs running on this host
These metrics are relevant to XenServer SRs, and will have "pool", "uuid", and "type" labels to distinguish them, as well as a "name_label" label for informational purposes.
default_storage: true if SR is a default SR for a pool, otherwise false
physical_pct_allocated: percent of physical_utilisation over physical_size
physical_size: total physical size of the repository (in bytes)
physical_utilisation: physical space currently utilised on this storage repository (in bytes). Note that for sparse disk formats, physical_utilisation may be less than virtual_allocation.
virtual_allocation: sum of virtual_sizes of all VDIs in this SR (in bytes)
A script exists to assist in generating a Docker container for xapi_exporter.
./build_container.sh xapi_exporter
- creates a build image to compile the go application
- copies the static binary out of the build image
- deletes the build image
- creates a run image using the static binary
The dual-image approach allows the final run image to be a tidy 35 MB in size.
The final image expects to be provided with /xapi_exporter/config.yml at run time.
docker run -v /local/xapi_exporter:/xapi_exporter -p 9290:9290 -t xapi_exporter