Service Documentation

Change process:

  1. Open ticket/set ticket to "in progress"

    1. If template change: Update template in lab first, test and then go to prod
    2. If settings change: Update settings in prod instance
  2. Dry run, if change might have a big impact implement on just a single device first or ask a colleague to verify diff as well
  3. Notify NOC if something might break
  4. Live run
  5. Check monitoring
  6. Update/close ticket

Firmware upgrade process:

  1. Open ticket/set ticket to "in progress"
  2. Verify in lab and document expected downtimes
  3. Upgrade isolated building/location first, wait 1 day
  4. Upgrade first half/partition of network
  5. Upgrade second half/partition of network
  6. Check monitoring
  7. Update/close ticket

Replace access switch:

  1. Save any special interface configs: curl -ks -H "Authorization: Bearer $JWT_AUTH_TOKEN" ${CNAASURL}/api/v1.0/device/eosaccess/interfaces
  2. Remove device from database
  3. ZTP new switch with same name
  4. Re-apply any interface configs ( migrate-interfaces.py compatible with nms v1.x to convert output json to input json , then curl ks -H "Authorization: Bearer $JWT_AUTH_TOKEN" ${CNAASURL}/api/v1.0/device/eosaccess/interfaces -X PUT -d outputfile.json -H "Content-Type: application/json" )
  5. Update serial etc in NI?
  6. Update monitoring etc if management IP changed

Replace dist switch (in core/dist configuration):

  1. Update mgmtdomain so it does not reference the switch you want to replace, instead set both device_a_id and device_b_id to the id of the device in the pair that you are not going to replace: curl ks -H "Authorization: Bearer $JWT_AUTH_TOKEN" ${CNAASURL}/api/v1.0/mgmtdomain/1 -X PUT -d '{"device_b_id": 8}'
  2. Delete the switch you want to replace with the API, factory_reset: false since it's not supported on DIST devices.
  3. Re-sync all core switches so they remove linknet config and add ZTP vlans
  4. Physically replace the switch and wait for the new switch to show up in DISCOVERED state
  5. Init the switch with the exact same hostname (so that git settings will be applied)
  6. Update the mgmtdomain so it references both switches in the pair again (revert of step 1)
  7. Update serial etc in NI?
  8. Update monitoring etc if management IP changed
  • No labels