Cray System Management Documentation > Upgrade CSM > scripts > sls > Upgrade SLS Offline from CSM 1.0.x to CSM 1.2

Upgrade SLS Offline from CSM 1.0.x to CSM 1.2

Abstract

Prior to updating SLS, at a minimum, answers to the following questions must be known:

Will user traffic (non-administrative) come in via the CAN or CHN, or is the site air-gapped?

By default when upgrading to CSM 1.2, non-administrative user traffic is migrated from the CAN to the CHN while minimizing changes. If instead it is desired that the non-administrative user traffic continue to come in via the CAN, or if the site is air-gapped, then this must be explicitly specified during the SLS update.
What is the internal VLAN and the site-routable IP subnet for the new CAN or CHN?
Is there a need to preserve any existing IP addresses during the CAN-to-CMN migration?
- One example is the external-dns IP address used for DNS lookups of system resources from site DNS servers. Changes to external-dns often require changes to site resources with requisite process and timeframes from other groups. For preserving external-dns IP addresses, the flag is --preserve-existing-subnet-for-cmn external-dns.
  
  WARNING: It is up to the administrator to compare pre-upgraded and post-upgraded SLS files for sanity. Specifically, in the case of preserving external-dns values, the administrator must ensure there are no site-networking changes that might result in NCN IP addresses overlapping during the upgrade process. This requires network subnetting expertise and “expert mode” (described below).
  
  If the external-dns IP address is changed, then the customizations.yaml site_to_system_lookups value must be updated to the new IP address. For instructions on how to do this, see Update customizations.yaml.
- A mutually exclusive example is the need to preserve all NCN IP addresses related to the old CAN while migrating the new CMN. This preservation is not often needed as the transition of NCN IP addresses for the CAN-to-CMN is automatically handled during the upgrade. The flag to preserve CAN-to-CMN NCN IP addresses is mutually exclusive with other preservations and the flag is --preserve-existing-subnet-for-cmn ncns.
- If no preservation flag is set, then the default behavior is to recalculate every IP address on the existing CAN while migrating to the CMN. The behavior in this case is to calculate the subnet sizes based on number of devices (with a bit of spare room), while maximizing IP address pool sizes for dynamic services.

Prerequisites

The latest CSM documentation RPM must be installed on the node where the procedure is being performed. See Check for Latest Documentation.

Procedure

(ncn#) Get a token.

export TOKEN=$(curl -s -k -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')

(ncn#) Extract SLS data to a file.

curl -k -H "Authorization: Bearer ${TOKEN}" https://api-gw-service-nmn.local/apis/sls/v1/dumpstate | jq -S . > sls_input_file.json

(ncn#) Upgrade SLS data.

Example 1: Upgrade, using the CHN as the system default route (will by default output to migrated_sls_file.json).

/usr/share/doc/csm/upgrade/scripts/sls/sls_updater_csm_1.2.py --sls-input-file sls_input_file.json \
    --bican-user-network-name CHN \
    --customer-highspeed-network 5 10.103.11.192/26

Example 2: Upgrade, using the CAN as the system default route, keep the generated CHN (for testing), and preserve the existing external-dns entry.

/usr/share/doc/csm/upgrade/scripts/sls/sls_updater_csm_1.2.py --sls-input-file sls_input_file.json \
    --bican-user-network-name CAN \
    --customer-highspeed-network 5 10.103.11.192/26 \
    --preserve-existing-subnet-for-cmn external-dns \
    --retain-unused-user-network

NOTE: A detailed review of the migrated/upgraded data is strongly recommended. Particularly, ensure that subnet reservations are correct to prevent any data loss.

(ncn#) Upload migrated SLS file to SLS service.

curl -H "Authorization: Bearer ${TOKEN}" -k -L -X POST 'https://api-gw-service-nmn.local/apis/sls/v1/loadstate' -F 'sls_dump=@migrated_sls_file.json'

SLS Updater help

(ncn#) For help and all options, run the following:

/usr/share/doc/csm/upgrade/scripts/sls/sls_updater_csm_1.2.py --help

Actions and order

This migration is performed offline for data security. The running SLS file is first dumped, then the migration script is run and a new, migrated output file is created.

Migrate switch naming (in order): leaf to leaf-bmc, and agg to leaf.
Remove api-gateway entries from HMLB subnets for CSM 1.2 security.
Remove kubeapi-vip reservations for all networks except NMN.
Create the new BICAN “toggle” network.
Migrate the existing CAN to CMN.
Create the CHN.
Convert IP addresses of the CAN.
Create MetalLB Pools and ASN entries on CMN and NMN.
Update uai_macvlan in NMN DHCP ranges and uai_macvlan VLAN.
Remove unused user networks (CAN or CHN) if requested (--retain-unused-user-network to keep).

Migrate switch names

Switch names change in CSM 1.2 and must be applied in the following order:

leaf switches become leaf-bmc switches.
agg switches become leaf switches.

This needs to be done in the order listed above.

Remove `api-gateway` / `istio-ingress-gateway` reservations from HMNLB subnets

For CSM 1.2, the API gateway no longer listens on the HMNLB MetalLB address pool. These aliases provided DNS records and are being removed.

Create the BICAN network “toggle”

New for CSM 1.2, the BICAN network ExtraProperties value of SystemDefaultRoute is used to point to the CAN, CHN, or CMN, and is used by utilities to systematically toggle routes.

Migrate existing CAN to new CMN

Using the existing CAN as a template, create the CMN. The same IP addresses will be preserved for NCNs (bootstrap_dhcp). A new network_hardware subnet will be created where the end of the previous bootstrap_dhcp subnet existed to contain switching hardware. MetalLB pools in the bootstrap_dhcp subnet will be shifted around to remain at the end of the new bootstrap subnet.

Create the CHN

With the original CAN as a template, the new CHN network will be created. IP addresses will come from the --customer-highspeed-network <vlan> <ipaddress> (or its defaults). This is be created by default, but can be removed (if not needed or desired) by using the --retain-unused-user-network flag.

Convert the IP addresses of the CAN

Since the original/existing CAN has been converted to the new CMN, the CAN must have new IP addresses. These are provided using the --customer-access-network <vlan> <ipaddress> (or its defaults). This CAN conversion will happen by default, but the new CAN may be removed (if not needed or desired) by using the --retain-unused-user-network flag.

Add BGP peering information to CMN and NMN

MetalLB and switches now obtain BGP peers using SLS data.

  --bgp-asn INTEGER RANGE         The autonomous system number for BGP router
                                  [default: 65533;64512<=x<=65534]
  --bgp-cmn-asn INTEGER RANGE     The autonomous system number for CMN BGP
                                  clients  [default: 65534;64512<=x<=65534]
  --bgp-nmn-asn INTEGER RANGE     The autonomous system number for NMN BGP
                                  clients  [default: 65533;64512<=x<=65534]

In CMN and NMN:

  "Type": "ethernet",
  "ExtraProperties": {
    "CIDR": "10.102.3.0/25",
    "MTU": 9000,
    "MyASN": 65536,
    "PeerASN": 65533,
    "Subnets":

Remove `kubeapi-vip` reservations for all networks except NMN

Self explanatory. This endpoint now exists only on the NMN.

Update `uai_macvlan` in NMN ranges and `uai_macvlan` VLAN

Self explanatory. Ranges are used for the addresses of UAIs.

Remove unused user networks (either CAN or CHN) if desired

By default the CAN will be removed if --bican-user-network-name CHN is specified, or the CHN will be removed if --bican-user-network-name CAN is specified. In order to prevent this, use the --retain-unused-user-network flag. Retention of the unused network is not normal behavior.

Generally production systems will NOT want to use this flag unless active toggling between CAN and CHN is required. This is not usual behavior.
Test/development systems may want to have all networks for testing purposes and might want to retain both user networks.

For technical details on the SLS update automation, see SLS Updater Technical Details.