Swap an HPE Cray EX liquid-cooled compute blade between two systems.
The two systems in this example are:
Source system - Cray EX TDS cabinet x9000 with a healthy EX425 blade (Windom dual-injection) in chassis 3, slot 0
Destination system - Cray EX cabinet x1005 with a defective EX425 blade (Windom dual-injection) in chassis 3, slot 0
Substitute the correct component names (xnames) or other parameters in the command examples that follow.
All the nodes in the blade must be specified using a comma-separated list. For example, EX425 compute blades include two node cards, each with two logical nodes (4 nodes).
The Slingshot fabric must be configured with the desired topology for both blades.
The System Layout Service (SLS) must have the desired HSN configuration.
The blade that is removed from the source system must be installed in the empty slot left by the blade removed from destination system, and vice-versa.
Check the status of the high-speed network (HSN) and record link status before the procedure.
Review the following command examples.
(ncn-mw#) The commands can be used to capture the required values from the HSM ethernetInterfaces table and write the values to a file.
The file then can be used to automate subsequent commands in this procedure, for example:
mkdir blade_swap_scripts; cd blade_swap_scripts
cat blade_query.sh
Example output:
#!/bin/bash
BLADE=$1
OUTFILE=$2
BLADE_DOT=$BLADE.
cray hsm inventory ethernetInterfaces list --format json | \
jq -c --arg BLADE "$BLADE_DOT" \
'map(select(.ComponentID|test($BLADE))) | map(select(.Description == "Node Maintenance Network")) | .[] | {xname: .ComponentID, ID: .ID,MAC: .MACAddress, IP: .IPAddresses[0].IPAddress,Desc: .Description}' > $OUTFILE
./blade_query.sh x1000c0s1 x1000c0s1.json
cat x1000c0s1.json
Example output:
{"xname":"x9000c3s1b0n0","ID":"0040a6836339","MAC":"00:40:a6:83:63:39","IP":"10.100.0.10","Desc":"Node Maintenance Network"}
{"xname":"x9000c3s1b0n1","ID":"0040a683633a","MAC":"00:40:a6:83:63:3a","IP":"10.100.0.98","Desc":"Node Maintenance Network"}
{"xname":"x9000c3s1b1n0","ID":"0040a68362e2","MAC":"00:40:a6:83:62:e2","IP":"10.100.0.123","Desc":"Node Maintenance Network"}
{"xname":"x9000c3s1b1n1","ID":"0040a68362e3","MAC":"00:40:a6:83:62:e3","IP":"10.100.0.122","Desc":"Node Maintenance Network"}
(ncn-mw#) To delete an ethernetInterfaces entry using curl:
for ID in $(cat x9000c3s1.json | jq -r '.ID'); do cray hsm inventory ethernetInterfaces delete $ID; done
(ncn-mw#) To insert an ethernetInterfaces entry using curl:
This requires an API token. See Retrieve an Authentication Token for more information.
while read PAYLOAD ; do
curl -H "Authorization: Bearer $TOKEN" -L -X POST 'https://api-gw-service-nmn.local/apis/smd/hsm/v2/Inventory/EthernetInterfaces' \
-H 'Content-Type: application/json' \
--data-raw "$(echo $PAYLOAD | jq -c '{ComponentID: .xname,Description: .Desc,MACAddress: .MAC,IPAddresses: [{IPAddress: .IP}]}')"
sleep 5
done < x9000c3s1.json
The blades must have the coolant drained and filled during the swap to minimize cross-contamination of cooling systems.
Using the work load manager (WLM), drain running jobs from the affected nodes on the blade. Refer to the vendor documentation for the WLM for more information.
(ncn-mw#) Use Boot Orchestration Services (BOS) to shut down the affected nodes in the source blade (in this example, x9000c3s0).
Specify the appropriate component name (xname) and BOS template for the node type in the following command.
BOS_TEMPLATE=cos-2.0.30-slurm-healthy-compute
cray bos v2 sessions create --template-name $BOS_TEMPLATE --operation shutdown --limit x9000c3s0b0n0,x9000c3s0b0n1,x9000c3s0b1n0,x9000c3s0b1n1
(ncn-mw#) Temporarily disable the Redfish endpoints for each compute node NodeBMC.
cray hsm inventory redfishEndpoints update --enabled false x9000c3s0b0 --id x9000c3s0b0 --hostname x9000c3s0b0
cray hsm inventory redfishEndpoints update --enabled false x9000c3s0b1 --id x9000c3s0b1 --hostname x9000c3s0b1
(ncn-mw#) Set the environment variable SLOT to the blade’s location.
SLOT="x9000c3s0"
(ncn-mw#) Clear the Redfish event subscriptions.
for BMC in $(cray hsm inventory redfishEndpoints list --type NodeBMC --format json | jq .RedfishEndpoints[].ID -r | grep ${SLOT}); do
/usr/share/doc/csm/scripts/operations/node_management/delete_bmc_subscriptions.py $BMC
done
Each BMC on the blade will have output like the following:
Clearing subscriptions from NodeBMC x3000c0s9b0
Retrieving BMC credentials from SCSD
Retrieving Redfish Event subscriptions from the BMC: https://x3000c0s9b0/redfish/v1/EventService/Subscriptions
Deleting event subscription: https://x3000c0s9b0/redfish/v1/EventService/Subscriptions/1
Successfully deleted https://x3000c0s9b0/redfish/v1/EventService/Subscriptions/1
(ncn-mw#) Remove the system specific settings from each node controller on the blade.
The two commands look similar, but note that the component names (xnames) in the URLs differ slightly, targeting each node controller on the blade.
curl -k -u root:PASSWORD -X POST -H \
'Content-Type: application/json' -d '{"ResetType":"StatefulReset"}' \
https://x9000c3s0b0/redfish/v1/Managers/BMC/Actions/Manager.Reset
curl -k -u root:PASSWORD -X POST -H \
'Content-Type: application/json' -d '{"ResetType":"StatefulReset"}' \
https://x9000c3s0b1/redfish/v1/Managers/BMC/Actions/Manager.Reset
Use Ctrl-C to return to the prompt if command does not return.
(ncn-mw#) Suspend the hms-discovery cronjob.
kubectl -n services patch cronjobs hms-discovery -p '{"spec" : {"suspend" : true }}'
(ncn-mw#) Verify that the hms-discovery cronjob has stopped.
That is, verify that ACTIVE = 0 and SUSPEND = True.
kubectl get cronjobs -n services hms-discovery
Example output:
NAME SCHEDULE SUSPEND ACTIVE LAST SCHEDULE AGE
hms-discovery */3 * * * * True 0 117s 15d
(ncn-mw#) Power off the chassis slot.
This examples powers off slot 0, chassis 3, in cabinet 9000.
cray power transition off x9000c3s0 --include children
(ncn-mw#) Disable the chassis slot.
Disabling the slot prevents hms-discovery from automatically powering on the slot. This example disables slot 0, chassis 3, in cabinet 9000.
cray hsm state components enabled update --enabled false x9000c3s0
IMPORTANT: Record the node management network (NMN) MAC and IP addresses for each node in the blade (labeled Node Maintenance Network). To prevent disruption in the
data virtualization service (DVS), these addresses must be maintained in the HSM when the blade is swapped and discovered.
The hardware management network MAC and IP addresses are assigned algorithmically and must not be deleted from the HSM.
(ncn-mw#) Query HSM to determine the ComponentID, MAC addresses, and IP addresses for each node in the blade.
The prerequisites show an example of how to gather HSM values and store them to a file.
cray hsm inventory ethernetInterfaces list --component-id x9000c3s0b0n0 --format json
Example output:
[
{
"ID": "0040a6836339",
"Description": "Node Maintenance Network",
"MACAddress": "00:40:a6:83:63:39",
"LastUpdate": "2021-04-09T21:51:04.662063Z",
"ComponentID": "x9000c3s0b0n0",
"Type": "Node",
"IPAddresses": [
{
"IPAddress": "10.100.0.10"
}
]
}
]
Record the following values for the blade:
"ComponentID": "x9000c3s0b0n0"
"MACAddress": "00:40:a6:83:63:39"
"IPAddress": "10.100.0.10"
Repeat the command to record the component IDs, MAC addresses, and IP addresses for the Node Maintenance Network for the other nodes in the blade.
(ncn-mw#) Delete the NMN MAC and IP addresses of each node in the blade from the HSM.
Do not delete the MAC and IP addresses for the node BMC.
cray hsm inventory ethernetInterfaces delete 0040a6836339
Repeat the preceding command for each node in the blade.
(ncn-mw#) Delete the Redfish endpoints for each node.
cray hsm inventory redfishEndpoints delete x9000c3s0b0
cray hsm inventory redfishEndpoints delete x9000c3s0b1
Use WLM to drain jobs from the affected nodes on the blade.
(ncn-mw#) Use BOS to shut down the affected nodes in the destination blade (in this example, x1005c3s0).
BOS_TEMPLATE=cos-2.0.30-slurm-healthy-compute
cray bos v2 sessions create --template-name $BOS_TEMPLATE --operation shutdown \
--limit x1005c3s0b0n0,x1005c3s0b0n1,x1005c3s0b1n0,x1005c3s0b1n1
(ncn-mw#) When nodes are Off, temporarily disable endpoint discovery service (MEDS) for the compute nodes.
Disabling chassis slot prevents hms-discovery from attempting to power them back on.
cray hsm inventory redfishEndpoints update --enabled false x1005c3s0b0 --id x1005c3s0b0 --hostname x1005c3s0b0
cray hsm inventory redfishEndpoints update --enabled false x1005c3s0b1 --id x1005c3s0b1 --hostname x1005c3s0b1
(ncn-mw#) Set the environment variable SLOT to the blade’s location.
SLOT="x9000c3s0"
(ncn-mw#) Clear the Redfish event subscriptions.
export TOKEN=$(curl -s -S -d grant_type=client_credentials \
-d client_id=admin-client \
-d client_secret=`kubectl get secrets admin-client-auth -o jsonpath='{.data.client-secret}' | base64 -d` \
https://api-gw-service-nmn.local/keycloak/realms/shasta/protocol/openid-connect/token | jq -r '.access_token')
for BMC in $(cray hsm inventory redfishEndpoints list --type NodeBMC --format json | jq .RedfishEndpoints[].ID -r | grep ${SLOT}); do
/usr/share/doc/csm/scripts/operations/node_management/delete_bmc_subscriptions.py $BMC
done
Each BMC on the blade will have output like the following:
Clearing subscriptions from NodeBMC x3000c0s9b0
Retrieving BMC credentials from SCSD
Retrieving Redfish Event subscriptions from the BMC: https://x3000c0s9b0/redfish/v1/EventService/Subscriptions
Deleting event subscription: https://x3000c0s9b0/redfish/v1/EventService/Subscriptions/1
Successfully deleted https://x3000c0s9b0/redfish/v1/EventService/Subscriptions/1
(ncn-mw#) Remove system-specific settings from each node controller on the blade.
The two commands look similar, but note that the component names (xnames) in the URLs differ slightly, targeting each node controller on the blade.
curl -k -u root:PASSWORD -X POST -H 'Content-Type: application/json' -d \
'{"ResetType": "StatefulReset"}' \
https://x1005c3s0b0/redfish/v1/Managers/BMC/Actions/Manager.Reset
curl -k -u root:PASSWORD -X POST -H 'Content-Type: application/json' -d \
'{"ResetType": "StatefulReset"}' \
https://x1005c3s0b1/redfish/v1/Managers/BMC/Actions/Manager.Reset
(ncn-mw#) Suspend the hms-discovery cronjob.
kubectl -n services patch cronjobs hms-discovery -p '{"spec": {"suspend": true }}'
(ncn-mw#) Destination: Verify that the hms-discovery cronjob has stopped.
That is, verify that ACTIVE = 0 and SUSPEND = True.
kubectl get cronjobs -n services hms-discovery
Example output:
NAME SCHEDULE SUSPEND ACTIVE LAST SCHEDULE AGE
hms-discovery */3 * * * * True 0 128s 15d
(ncn-mw#) Destination: Power off the chassis slot.
This example powers off slot 0 in chassis 3 of cabinet 1005.
cray power transition off x1005c3s0 --include children
(ncn-mw#) Disable the chassis slot.
This example disables slot 0, chassis 3, in cabinet 1005.
cray hsm state components enabled update --enabled false x1005c3s0
IMPORTANT: Record the component ID, MAC addresses, and IP addresses for each node in the blade in the destination system. To prevent disruption in the data virtualization service (DVS), these addresses must be maintained in the HSM when the replacement blade is swapped and discovered.
The hardware management network NIC MAC addresses for liquid-cooled blades are assigned algorithmically and must not be deleted from the HSM.
(ncn-mw#) Query HSM to determine the component ID, MAC addresses, and IP addresses associated with nodes in the destination blade.
The prerequisites show an example of how to gather HSM values and store them to a file.
cray hsm inventory ethernetInterfaces list \
--component-id XNAME --format json
Example output:
[
{
"ID": "0040a6836399",
"Description": "Node Maintenance Network",
"MACAddress": "00:40:a6:83:63:99",
"LastUpdate": "2021-04-09T21:51:04.662063Z",
"ComponentID": "x1005c3s0b0n0",
"Type": "Node",
"IPAddresses": [
{
"IPAddress": "10.100.0.123"
}
]
}
]
(ncn-mw#) Record the following Node Maintenance Network values for each node in the blade.
"ComponentID": "x1005c3s0b0n0"
"MACAddress": "00:40:a6:83:63:99"
"IPAddress": "10.10.0.123"
(ncn-mw#) Delete the node NIC MAC and IP addresses from the HSM ethernetInterfaces table for each node.
cray hsm inventory ethernetInterfaces delete 0040a6836399
Repeat the preceding command for each node in the blade.
(ncn-mw#) Delete the Redfish endpoints for each node.
cray hsm inventory redfishEndpoints delete x1005c3s0b0
cray hsm inventory redfishEndpoints delete x1005c3s0b1
Remove the blade from destination system install it in a storage cart.
Install the blade from the source system into the destination system.
(ncn-mw#) Obtain an authentication token to access the API gateway.
In the example below, replace myuser, mypass, and shasta in the curl command with site-specific values. Note
the value of access_token. Review Retrieve an Authentication Token for more information. The example is a
script to secure a token and set it to the variable TOKEN.
export TOKEN=$(curl -s -d grant_type=password -d client_id=shasta -d \
username=USERNAME -d password=PASSWORD \
https://api-gw-service-nmn.local/keycloak/realms/shasta/protocol/openid-connect/token | jq -r '.access_token')
echo $TOKEN
(ncn-mw#) Stop Kea from automatically adding MAC address entries to the HSM ethernetInterfaces table.
Kea automatically adds entries to the HSM ethernetInterfaces table when a DHCP lease is provided (about every 5 minutes). To prevent Kea from automatically adding these entries,
perform the following steps:
Create an eth_interfaces file that contains the interface IDs for the Node Maintenance Network entries for the destination system.
When repeating this procedure for the source system, use the interface IDs for the source system.
cat eth_interfaces
Example output:
0040a6836339
0040a683633a
0040a68362e2
0040a68362e3
Run the following commands in succession to remove the interfaces.
Deleting the cray-dhcp-kea pod prevents the interfaces from being re-created.
kubectl delete -n services pod $(kubectl get pods -n services | grep cray-dhcp-kea- | awk '{ print $2 }')
for ETH in $(cat eth_interfaces); do cray hsm inventory ethernetInterfaces delete $ETH --format json ; done
(ncn-mw#) Add the MAC address, IP address, and Node Maintenance Network description to the interfaces.
The component ID and IP address must be the values recorded from the destination blade, and the MAC address must be the value recorded from the source blade.
"ComponentID": "x1005c3s0b0n0"
"MACAddress": "00:40:a6:83:63:99"
"IPAddress": "10.10.0.123"
MAC=SOURCESYS_MAC_ADDRESS
IP_ADDRESS=DESTSYS_IP_ADDRESS
XNAME=DESTSYS_XNAME
curl -H "Authorization: Bearer ${TOKEN}" -L -X POST 'https://api-gw-service-nmn.local/apis/smd/hsm/v2/Inventory/EthernetInterfaces' \
-H 'Content-Type: application/json' --data-raw "{
\"Description\": \"Node Maintenance Network\",
\"MACAddress\": \"$MAC\",
\"IPAddresses\": [
{
\"IPAddress\": \"$IP_ADDRESS\"
}
],
\"ComponentID\": \"$XNAME\"
}"
When repeating this procedure for the source system, the component ID and IP address must be the values recorded from the source system, and the MAC address must be the value recorded from the blade in the destination system.
MAC=DESTSYS_MAC_ADDRESS
IP_ADDRESS=SOURCESYS_IP_ADDRESS
XNAME=SOURCESYS_XNAME
To change or correct a curl command that has been entered, use a PATCH request. For example:
curl -k -H "Authorization: Bearer $TOKEN" -L -X PATCH \
'https://api-gw-service-nmn.local/apis/smd/hsm/v2/Inventory/EthernetInterfaces/0040a68350a4' -H 'Content-Type: application/json' \
--data-raw '{"MACAddress":"xx:xx:xx:xx:xx:xx","IPAddresses":[{"IPAddress":"10.xxx.xxx.xxx"}],"ComponentID":"XNAME"}'
(ncn-mw#) Restart Kea.
kubectl delete pods -n services -l app.kubernetes.io/name=cray-dhcp-kea
Repeat the preceding step for each node in the blade.
(ncn-mw#) Enable the chassis slot.
The example enables slot 0, chassis 3, in cabinet 1005.
cray hsm state components enabled update --enabled true x1005c3s0
(ncn-mw#) Power on the chassis slot.
The example powers on slot 0, chassis 3, in cabinet 1005.
cray power transition off x1005c3s0 --include children
(ncn-mw#) Verify that the hms-discovery cronjob is not suspended in Kubernetes.
That is, verify that ACTIVE = 1 and SUSPEND = False.
kubectl -n services patch cronjobs hms-discovery \
-p '{"spec" : {"suspend" : false }}'
kubectl get cronjobs.batch -n services hms-discovery
Example output:
NAME SCHEDULE SUSPEND ACTIVE LAST SCHEDULE AGE
hms-discovery */3 * * * * False 1 41s 33d
kubectl get pods -Ao wide | grep hms-discovery
kubectl -n services logs hms-discovery-1600117560-5w95d \
hms-discovery | grep "Mountain discovery finished" | jq '.discoveredXnames'
Example output:
[
"x1005c3s0b0"
]
Wait for three minutes for the blade to power on and the node controllers (BMCs) to be discovered.
(ncn-mw#) To verify that the BMCs have been discovered by the HSM, run this command for each BMC in the blade.
cray hsm inventory redfishEndpoints describe XNAME --format json
Example output:
{
"ID": "x1005c3s0b0",
"Type": "NodeBMC",
"Hostname": "x1005c3s0b0",
"Domain": "",
"FQDN": "x1005c3s0b0",
"Enabled": true,
"User": "root",
"Password": "",
"MACAddr": "02:03:E8:00:31:00",
"RediscoverOnUpdate": true,
"DiscoveryInfo": {
"LastDiscoveryAttempt": "2021-06-10T18:01:59.920850Z",
"LastDiscoveryStatus": "DiscoverOK",
"RedfishVersion": "1.2.0"
}
}
LastDiscoveryStatus displays as DiscoverOK, the node BMC has been successfully discovered.DiscoveryStarted, then the BMC is currently being inventoried by HSM.HTTPsGetFailed or ChildVerificationFailed, then an error has
occurred during the discovery process.(ncn-mw#) Optional: Force rediscovery of the components in the chassis.
This example shows cabinet 1005, chassis 3.
cray hsm inventory discover create --xnames x1005c3b0
(ncn-mw#) Optional: Verify that discovery has completed.
That is, verify that LastDiscoveryStatus = DiscoverOK.
cray hsm inventory redfishEndpoints describe x1005c3 --format json
Example output:
{
"ID": "x1005c3",
"Type": "ChassisBMC",
"Hostname": "x1005c3",
"Domain": "",
"FQDN": "x1005c3",
"Enabled": true,
"User": "root",
"Password": "",
"MACAddr": "02:03:ED:03:00:00",
"RediscoverOnUpdate": true,
"DiscoveryInfo": {
"LastDiscoveryAttempt": "2020-09-03T19:03:47.989621Z",
"LastDiscoveryStatus": "DiscoverOK",
"RedfishVersion": "1.2.0"
}
}
(ncn-mw#) Enable the nodes in the HSM database.
cray hsm state components bulkEnabled update --enabled true \
--component-ids x1005c3s0b0n0,x1005c3s0b0n1,x1005c3s0b1n0,x1005c3s0b1n1
(ncn-mw#) Verify that the nodes are enabled in the HSM.
cray hsm state components query create \
--component-ids x1005c3s0b0n0,x1005c3s0b0n1,x1005c3s0b1n0,x1005c3s0b1n1 --format toml
Partial example output:
[[Components]]
ID = x1005c3s0b0n0
Type = "Node"
Enabled = true
State = "Off"
[[Components]]
ID = x1005c3s0b1n1
Type = "Node"
Enabled = true
State = "Off"
(ncn-mw#) Use boot orchestration to power on and boot the nodes.
Specify the appropriate BOS template for the node type.
BOS_TEMPLATE=cos-2.0.30-slurm-healthy-compute
cray bos v2 sessions create --template-name $BOS_TEMPLATE \
--operation reboot --limit x1005c3s0b0n0,x1005c3s0b0n1,x1005c3s0b1n0,x1005c3s0b1n1
Verify that the correct firmware versions for node BIOS, node controller (nC), NIC mezzanine card (NMC), GPUs, and so on.
(ncn-mw#) If necessary, update the firmware.
Review the FAS Admin Procedures and Update Firmware with FAS procedure.
cray fas actions create CUSTOM_DEVICE_PARAMETERS.json
There should be one or more cray-cps pods.
(ncn-mw#) Verify that the cray-cps pods on worker nodes are Running.
kubectl get pods -Ao wide | grep cps
Example output:
services cray-cps-75cffc4b94-j9qzf 2/2 Running 0 42h 10.40.0.57 ncn-w001
(ncn-w#) SSH to each worker node running CPS/DVS, and run dmesg -T to ensure that there are no recurring "DVS: merge_one" error messages as shown.
The error messages indicate that DVS is detecting an IP address change for one of the client nodes.
dmesg -T | grep "DVS: merge_one"
[Tue Jul 21 13:09:54 2020] DVS: merge_one#351: New node map entry does not match the existing entry
[Tue Jul 21 13:09:54 2020] DVS: merge_one#353: nid: 8 -> 8
[Tue Jul 21 13:09:54 2020] DVS: merge_one#355: name: 'x3000c0s19b1n0' -> 'x3000c0s19b1n0'
[Tue Jul 21 13:09:54 2020] DVS: merge_one#357: address: '10.252.0.26@tcp99' -> '10.252.0.33@tcp99'
[Tue Jul 21 13:09:54 2020] DVS: merge_one#358: Ignoring.
Make sure the Configuration Framework Service (CFS) finished successfully. Review HPE Cray EX DVS Administration Guide 1.4.1 S-8004.
(nid#) SSH to the node and check each DVS mount.
mount | grep dvs | head -1
Example output:
/var/lib/cps-local/0dbb42538e05485de6f433a28c19e200 on /var/opt/cray/gpu/nvidia-squashfs-21.3 type dvs (ro,relatime,blksize=524288,statsfile=/sys/kernel/debug/dvs/mounts/1/stats,attrcache_timeout=14400,cache,nodatasync,noclosesync,retry,failover,userenv,noclusterfs,killprocess,noatomic,nodeferopens,no_distribute_create_ops,no_ro_cache,loadbalance,maxnodes=1,nnodes=6,nomagic,hash_on_nid,hash=modulo,nodefile=/sys/kernel/debug/dvs/mounts/1/nodenames,nodename=x3000c0s6b0n0:x3000c0s5b0n0:x3000c0s4b0n0:x3000c0s9b0n0:x3000c0s8b0n0:x3000c0s7b0n0)
ls /var/opt/cray/gpu/nvidia-squashfs-21.3
Example output:
rootfs
(ncn-mw#) Determine the pod name for the Slingshot fabric manager pod and check the status of the fabric.
kubectl exec -it -n services \
$(kubectl get pods --all-namespaces |grep slingshot | awk '{print $2}') \
-- fmn_status
(ncn-mw#) Check for duplicate IP address entries in the State Management Database (SMD).
Duplicate entries will cause DNS operations to fail.
ssh uan01
Example output:
ssh: Could not resolve hostname uan01: Temporary failure in name resolution
ssh x3000c0s14b0n0
Example output:
ssh: Could not resolve hostname x3000c0s14b0n0: Temporary failure in name resolution
ssh x1000c1s1b0n1
Example output:
ssh: Could not resolve hostname x1000c1s1b0n1: Temporary failure in name resolution
The Kea configuration error will display a message similar to the message below. This message indicates a duplicate IP address (10.100.0.105) in the SMD:
Config reload failed
[{'result': 1, 'text': "Config reload failed: configuration error using file '/usr/local/kea/cray-dhcp-kea-dhcp4.conf':
failed to add new host using the HW address '00:40:a6:83:50:a4 and DUID '(null)' to the IPv4 subnet id '0' for the address 10.100.0.105: There's already a reservation for this address"}]
(ncn-mw#) Check for active DHCP leases.
If there are no DHCP leases, then there is a configuration error.
This requires an API token. See Retrieve an Authentication Token for more information.
curl -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
Example output in the case of a configuration error:
[
{
"arguments": {
"leases": []
},
"result": 3,
"text": "0 IPv4 lease(s) found."
}
]
(ncn-mw#) Delete any duplicate entries.
If there are duplicate entries in the SMD as a result of the swap procedure (10.100.0.105 in this example), then delete the duplicate entry.
Show the EthernetInterfaces for the duplicate IP address:
cray hsm inventory ethernetInterfaces list --ip-address 10.100.0.105 --format json | jq
Example output:
[
{
"ID": "0040a68350a4",
"Description": "Node Maintenance Network",
"MACAddress": "00:40:a6:83:50:a4",
"IPAddress": "10.100.0.105",
"LastUpdate": "2021-08-24T20:24:23.214023Z",
"ComponentID": "x1000c7s7b0n1",
"Type": "Node"
},
{
"ID": "0040a683639a",
"Description": "Node Maintenance Network",
"MACAddress": "00:40:a6:83:63:9a",
"IPAddress": "10.100.0.105",
"LastUpdate": "2021-08-27T19:15:53.697459Z",
"ComponentID": "x1000c7s7b0n1",
"Type": "Node"
}
]
Delete the older entry.
cray hsm inventory ethernetInterfaces delete 0040a68350a4
(ncn-mw#) Check DNS.
nslookup 10.252.1.29
Example output:
1.1.252.10.in-addr.arpa name = uan01.
1.1.252.10.in-addr.arpa name = uan01.local.
1.1.252.10.in-addr.arpa name = x3000c0s14b0n0.
1.1.252.10.in-addr.arpa name = x3000c0s14b0n0.local.
1.1.252.10.in-addr.arpa name = uan01-nmn.
1.1.252.10.in-addr.arpa name = uan01-nmn.local.
nslookup uan01
Example output:
Server: 10.92.100.225
Address: 10.92.100.225#53
Name: uan01
Address: 10.252.1.29
nslookup x3000c0s14b0n0
Example output:
Server: 10.92.100.225
Address: 10.92.100.225#53
Name: x3000c0s14b0n0
Address: 10.252.1.29
(ncn-mw#) Check SSH.
ssh x3000c0s14b0n0
Example output:
The authenticity of host 'x3000c0s14b0n0 (10.252.1.29)' can't be established.
ECDSA key fingerprint is SHA256:wttHXF5CaJcQGPTIq4zWp0whx3JTwT/tpx1dJNyyXkA.
Are you sure you want to continue connecting (yes/no/[fingerprint])? yes
Warning: Permanently added 'x3000c0s14b0n0' (ECDSA) to the list of known hosts.
Last login: Tue Aug 31 10:45:49 2021 from 10.252.1.9
To minimize cross-contamination of cooling systems, drain the coolant from the blade removed from destination system and fill with fresh coolant.
Review the HPE Cray EX Coolant Service Procedures H-6199. If using the hand pump, review procedures in the HPE Cray EX Hand Pump User Guide H-6200 (HPE Support).
Install the blade from the destination system into source system.
Review the Remove a Compute Blade Using the Lift procedure in HPE Cray EX Hardware Replacement Procedures H-6173 for detailed instructions for replacing liquid-cooled blades (HPE Support).
Power on the nodes in the source system by repeating the steps starting in the Bring up the blade in the destination system section, and going up through the Check DNS section.