# Verify descriptor matches flat extent size DESCRIPTOR_SIZE=$(grep "RW" $vmdk | awk 'print $2') FLAT_SIZE=$(stat -c %s $vmdk%-flat.vmdk.vmdk 2>/dev/null)
I pull the manifest. The VMDK is 2TB thin-provisioned. The VM won't boot. The backup ran six hours ago, but the differential is useless because the block map is scrambled.
"Server is slow. Throwing I/O errors. Tried to restart; now it won't boot."
# Check for zeroed CID (Corruption flag) CID=$(grep "CID=" $vmdk | cut -d'=' -f2) if [ "$CID" == "fffffffe" ]; then echo "ALERT: $vmdk has invalid CID. Possible corruption." >> $LOG_FILE /usr/bin/vmkfstools -x check $vmdk >> $LOG_FILE fi
Upon attempting to power on the VM, vSphere returned the dreaded: "Failed to power on VM. The specified virtual disk needs repair (Corrupt disk)." Attempting to browse the datastore revealed the file server01_2.vmdk (the 200GB delta disk) present, but the system could not read its metadata. Part 1: The Post-Mortem Analysis What is a VMDK? A VMDK is not a single file, but a logical container. It consists of a small descriptor file (text) and a large extent file (raw data). Corruption often strikes the descriptor or the filesystem journal within the extent.
[SUCCESS] Grain table rebuilt. 1,892,471 sectors recovered. [WARNING] 204 sectors unrecoverable (log files, temporary data). The SQL database performs a recovery on mount. I lose 15 minutes of transaction log data. The business accepts the loss. To prevent this, implement a VMDK health check cron job on the ESXi host:
