The Domain Name Service (DNS) is part of an integrated infrastructure set designed to provide dynamic host discovery, addressing, and naming. There are several different place to look for troubleshooting as DNS interacts with Dynamic Host Configuration Protocol (DHCP), the Hardware Management Service (HMS), the System Layout Service (SLS), and the State Manager Daemon (SMD).
The information below describes what to check when experiencing issues with DNS.
It is important to verify if a hostname is correct. The values in the networks.yml or networks_derived.yml files are sometimes inaccurate.
The formats show below are valid hostnames:
<xname\><xname\>.local<xname\>-mgmt<xname\>-mgmt.localnid<nid\_number\>-nmnnid<nid\_number\>-nmn.localAdditional steps are needed if a hostname or component name (xname) is either listed incorrectly or not listed at all in the networks.yml or networks_derived.yml files.
In that case, the following actions must be taken:
Use the dig or nslookup commands directly against the Unbound resolver. A host is correctly in DNS if the response from the dig command includes the following:
ANSWER SECTION value exists with a valid hostname and IP address.QUERY value exists that has the status: NOERROR message.ncn# dig HOSTNAME @10.92.100.225
Example output:
; <<>> DiG 9.11.2 <<>> x3000c0r41b0 @10.92.100.225
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 57196
;; flags: qr aa rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 1
;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 4096
;; QUESTION SECTION:
;x3000c0r41b0. IN A
;; ANSWER SECTION:
x3000c0r41b0. 3600 IN A 10.254.127.200
;; Query time: 0 msec
;; SERVER: 10.92.100.225#53(10.92.100.225)
;; WHEN: Fri Jul 17 18:49:48 UTC 2020
;; MSG SIZE rcvd: 57
If either of the commands fail to meet the two conditions mentioned above, then collect the logs to troubleshoot.
If there no record in the Unbound pod, that is also an indication that the host is not in DNS.
ncn-mw# kubectl -n services get cm cray-dns-unbound -o jsonpath='{.binaryData.records\.json\.gz}' |
base64 --decode | gzip -dc | jq -c '.[]' | grep XNAME
Example output excerpt:
{"hostname": "x1003c7s7b0", "ip-address": "10.104.12.191"}
cray-dns-unbound logs for errorsUse the following command to check the logs. Any logs with a message saying ERROR or Exception are an indication that the Unbound service is not healthy.
ncn-mw# kubectl logs -n services -l app.kubernetes.io/instance=cray-dns-unbound -c cray-dns-unbound
Example output excerpt:
[1596224129] unbound[8:0] debug: using localzone health.check.unbound. transparent
[1596224129] unbound[8:0] debug: using localzone health.check.unbound. transparent
[1596224135] unbound[8:0] debug: using localzone health.check.unbound. transparent
[1596224135] unbound[8:0] debug: using localzone health.check.unbound. transparent
To view the DNS Helper logs:
ncn-mw# kubectl logs -n services pod/$(kubectl get -n services pods |
grep unbound | tail -n 1 | cut -f 1 -d ' ') -c manager | tail -n4
Example output:
uid: bc1e8b7f-39e2-49e5-b586-2028953d2940
Comparing new and existing DNS records.
No differences found. Skipping DNS update
Log in to the spine switches and verify that MetalLB is peering to the spines via BGP.
Check both spines if they are available and powered up. All worker nodes should be peered with the spine BGP.
sw-spine# show ip bgp neighbors
Example output:
BGP neighbor: 10.252.0.4, remote AS: 65533, link: internal:
Route-map (in/out) : rm-ncn-w001
BGP version : 4
Configured hold time in seconds : 180
keepalive interval in seconds (configured) : 60
keepalive interval in seconds (established with peer): 30
Minimum holdtime from neighbor in seconds : 90BGP neighbor: 10.252.0.5, remote AS: 65533, link: internal:
Route-map (in/out) : rm-ncn-w002
BGP version : 4
Configured hold time in seconds : 180
keepalive interval in seconds (configured) : 60
keepalive interval in seconds (established with peer): 30
Minimum holdtime from neighbor in seconds : 90BGP neighbor: 10.252.0.6, remote AS: 65533, link: internal:
Route-map (in/out) : rm-ncn-w003
BGP version : 4
Configured hold time in seconds : 180
keepalive interval in seconds (configured) : 60
keepalive interval in seconds (established with peer): 30
Minimum holdtime from neighbor in seconds : 90
Confirm that routes to Kea (10.92.100.222) via all the NCN worker nodes are available:
sw-spine# show ip route 10.92.100.222
Example output:
Flags:
F: Failed to install in H/W
B: BFD protected (static route)
i: BFD session initializing (static route)
x: protecting BFD session failed (static route)
c: consistent hashing
p: partial programming in H/W
VRF Name default:
------------------------------------------------------------------------------------------------------
Destination Mask Flag Gateway Interface Source AD/M
------------------------------------------------------------------------------------------------------
default 0.0.0.0 c 10.102.255.9 eth1/16 static 1/1
10.92.100.222 255.255.255.255 c 10.252.0.4 vlan2 bgp 200/0
c 10.252.0.5 vlan2 bgp 200/0
c 10.252.0.6 vlan2 bgp 200/0
tcpdumpVerify that the NCN is receiving DNS queries. On an NCN worker or manager with kubectl installed, run the following command:
ncn-mw# tcpdump -envli bond0.nmn0 port 53
ping and ssh commands fail for hosts in DNSIf the IP address returned by the ping command is different than the IP address returned by the dig command, then restart nscd on the impacted node. This is done with the following command:
ncn# systemctl restart nscd.service
Attempt to ping or ssh to the IP address that was experiencing issues after restarting nscd.
Search for a DHCP lease by checking active leases for the service:
ncn# curl -s -k -H "Authorization: Bearer ${TOKEN}" -X POST -H \
"Content-Type: application/json" \-d '{ "command": "lease4-get-all", "service": \
[ "dhcp4" ] }' https://api-gw-service-nmn.local/apis/dhcp-kea | jq
For example:
ncn# curl -s -k -H "Authorization: Bearer ${TOKEN}" -X \
POST -H "Content-Type: application/json" \-d '{ "command": "lease4-get-all", "service": \
[ "dhcp4" ] }' https://api-gw-service-nmn.local/apis/dhcp-kea \
| jq | grep x3000c0s19b4 -A 6 -B 4
Example output:
{
"cltt": 1597777241,
"fqdn-fwd": true,
"fqdn-rev": true,
"hostname": "x3000c0s19b4",
"hw-address": "a4:bf:01:3e:d2:94",
"ip-address": "10.254.127.205",
"state": 0,
"subnet-id": 1,
"valid-lft": 300
}
If there is not a DHCP lease found, then: