Ever thought about automated RMAs? What could go right ?
Arno Dubois from Criteo presents issues at operating RMAs on a mid-size infrastructure and proposes an API-oriented solution to solve this.
DCs 2 DCs 1 DC 1 DC Interactions with contractors? Interactions with vendors? SoX: nobody else but us + contractor can get into the cage We are 3 for 32k servers, ~110 tickets/month 1 DC
Intervention ticket created Vendor Order the part Dispatch the part Analyze the problem Receive part Replace part Send back the old part Acknowledge part Verify fix 5-10mins/part
Intervention ticket created Vendor Order the part Dispatch the part Analyze the problem Receive part Replace part Send back the old part Acknowledge part Verify fix 5-10mins/part OUR CONTRACTOR 3 TIMES A DAY 16 HOURS PER MONTH
Intervention ticket created Vendor Order the part Dispatch the part Analyze the problem Receive part Replace part Send back the old part Acknowledge part Verify fix 5-10mins/part OUR CONTRACTOR 3 TIMES A DAY 16 HOURS PER MONTH
the agnostic SKU approach (looking at you, HDDs and SSDs) • Includes details if required by vendor Source: 8 PUT /rma { "part_id": "HU_MB_I4512365QSD", // Exact part (Unique amongst vendors!) "agnostic_part_id": "SHU_MB_IJHS75BDS56", // Agnostic part (also unique) "exact_part": true, // Do you accept another part if the exact one is not available? "server_identifier": "MSXPA91294", // Could be a serial, a service tag, whatever, "details?": { "description": "Server does not boot, blocks at POST", "steps_taken": "Cleared BIOS settings, reset BMC", "error_codes": ["0x00000001", "0x00000002"], "additional_info?": "Server was working fine until a power outage", } }
Intervention ticket created Vendor Order the part Dispatch the part Analyze the problem Receive part Replace part Send back the old part Acknowledge part Verify fix 5-10mins/part
ticket created Vendor Order the part Dispatch the part Analyze the problem Receive part Replace part Send back the old part Acknowledge part Verify fix 0mins/part