exporter cannot parse network_ip_availabilities_total metrics

[rrraditya]

Openstack-exporter version used:

openstack-exporter, version 1.0.0 (branch: HEAD, revision: e6224ef)
build user: root@fd8bce58a223
build date: 20200327-21:44:22
go version: go1.13.9

What happened:
If there's a subnet with IPv6 with high allocation pools in the openstack cluster, the exporter fail to parse the value.
error was given :

Oct 7 10:42:23 prom01 openstack-exporter: time="2021-10-07T10:42:23+07:00" level=info msg="Collecting metrics for exporter: openstack_neutron, metric: network_ip_availabilities_total" source="exporter.go:88"
Oct 7 10:42:23 01prom01 openstack-exporter: time="2021-10-07T10:42:23+07:00" level=error msg="math/big: cannot unmarshal "1.1805916207174113e+21" into a *big.Int" source="exporter.go:96"

What you expected to happen:
no error found from openstack-exporter when there's IPv6 subnet in the openstack cluster

How to reproduce it:

• create IPv6 subnet with 2001:db8::/58 or 2001:db8::/56
• set gateway to 2001:db8::1
• set allocation pools 2001:db8::2,2001:db8:0:3f:ffff:ffff:ffff:ffff
• Validate openstack-exporter logs, look for error when it tries to collect network_ip_availabilities_total metrics

[niedbalski]

Hey @alexeymyltsev is this something you have experienced with the network_ip_availabilities_total metric?

[rrraditya]

hi @niedbalski
just adding our finding that this issue also occured in openstack-exporter latest release which is v1.4.0

/usr/local/bin/openstack-exporter-1.4 --version
openstack-exporter, version 1.4.0 (branch: HEAD, revision: adab1c3)
build user: root@7c71a7ee40cd
build date: 20210527-04:07:51
go version: go1.13.15

ps -ef | grep 53428
root 53428 1 3 02:41 ? 00:00:18 /usr/local/bin/openstack-exporter-1.4 --web.listen-address=:9181 --os-client-config="/etc/openstack/clouds.yaml" nc-region-1

Oct 08 02:47:49 prom02 openstack-exporter-1.4[53428]: time="2021-10-08T02:47:49Z" level=info msg="Collecting metrics for exporter: openstack_neutron, metric: network_ip_availabilities_total" source="exporter.go:96"
Oct 08 02:47:49 prom02 openstack-exporter-1.4[53428]: time="2021-10-08T02:47:49Z" level=error msg="Failed to collect metric for exporter: neutron, error: failed to collect metric: network_ip_availabilities_total, error: math/big: cannot unmarshal "4.722366482869645e+21" into a *big.Int" source="exporter.go:122"

hope this information helps.

Thank you.

Regards,
Reqi

[alexeymyltsev]

Hey @alexeymyltsev is this something you have experienced with the network_ip_availabilities_total metric?

Hi @niedbalski ! Unfortunately I didn't face with problem like that and we do no use IPv6, but I can try to reproduce this problem on our clusters.

[alexeymyltsev]

I have done several tests and I just find that cause of problem in gophercloud.
With CURL I got:
{"network_ip_availabilities": [{"network_id": "484e4bc5-902f-4637-bd90-38d0befafc9e", "network_name": "test-ipv6", "tenant_id": "281707607f27476e955432b1ed0e1f87", "project_id": "281707607f27476e955432b1ed0e1f87", "subnet_ip_availability": [{"subnet_id": "9464a1d1-be98-4408-89ee-5c125249b1b5", "ip_version": 6, "cidr": "2001:db8::/56", "subnet_name": "2001:db8::/56", "used_ips": 3, "total_ips": 4722366482869645213694}], "used_ips": 3, "total_ips": 4722366482869645213694}]}

But if I use gophercloud, I got:
{{{map[network_ip_availability:map[network_id:484e4bc5-902f-4637-bd90-38d0befafc9e network_name:test-ipv6 project_id:281707607f27476e955432b1ed0e1f87 subnet_ip_availability:[map[cidr:2001:db8::/56 ip_version:6 subnet_id:9464a1d1-be98-4408-89ee-5c125249b1b5 subnet_name:2001:db8::/56 total_ips:4.722366482869645e+21 used_ips:3]] tenant_id:281707607f27476e955432b1ed0e1f87 total_ips:4.722366482869645e+21 used_ips:3]] map[Content-Length:[472] Content-Type:[application/json] Date:[Sat, 09 Oct 2021 19:37:19 GMT] Vary:[Accept-Encoding] X-Openstack-Request-Id:[req-36b4d6c3-6db6-4493-ae92-177442d72d27]]

Unfortunately I have not find yet how this problem could be fixed.

Looks like this issue linked with https://tools.ietf.org/html/rfc7159#section-6

[rrraditya]

hi @alexeymyltsev could you share how we can do the curl to get specific network_ip_availabilities for specific network_id? so we can test it too from our side.

[alexeymyltsev]

@rrraditya
curl -s -H "X-Auth-Token: $OS_TOKEN" https://api.example.com:9696/v2.0/network-ip-availabilities\?network_id\=484e4bc5-902f-4637-bd90-38d0befafc9e

[rrraditya]

Thank you @alexeymyltsev
we'll test it on our side.

meanwhile, about this case, so there's still no plan whether this will be fixed from gophercloud side?
i mean, does this case will be marked as bug and will be fixed or else?

Thank you.

[alexeymyltsev]

Thank you @alexeymyltsev we'll test it on our side.

meanwhile, about this case, so there's still no plan whether this will be fixed from gophercloud side? i mean, does this case will be marked as bug and will be fixed or else?

Thank you.

I am planing create issue to gophercloud project.
I will add here link to issue for gothercloud.

[alexeymyltsev]

Issue for gophercloud
gophercloud/gophercloud#2233

[rrraditya]

thank you @alexeymyltsev for the reference.
We'll keep an eye for updates on that ticket.

[curveds]

Probably will help someone to avoid math/big error (workaround):

func ListNetworkIPAvailabilities(exporter *BaseOpenStackExporter, ch chan<- prometheus.Metric) error {
	var allNetworkIPAvailabilities []networkipavailabilities.NetworkIPAvailability

	allPagesNetworkIPAvailabilities, err := networkipavailabilities.List(exporter.Client, networkipavailabilities.ListOpts{
		IPVersion: "4", // Gather metrics only for IPv4
	}).AllPages()
	if err != nil {
		return err
	}

IPv6 subnets are big enough anyway.

[NamNH2393]

Hi team
This issue not resolve after many release version
Does we have any plan for this !?

Hope to receive feedback from team
Thanks team