Resolving a problem with failure to boot

Edit online

An IBM® Service Support Representative (SSR) can use this procedure if a node canister fails to boot when powered on.

Before you begin

If the Spectrum Virtualize software is not running, then the node canister status and battery status LEDs are off. The service interfaces such as the service IP, technician port, and satask.txt on a USB flash drive do not work.
Note: If the canister services software on the PCIe switch chip finds that the microprocessors might not start, the node canister fault LED might be on with the node canister status LED off.

If the Spectrum Virtualize software is running, then the node canister fault LED might be on, and the node canister status LED might blink. The node canister error code and error data can be seen in the Service Assistant GUI, by connecting to the technician port, or by using the other service interfaces. Look up the error code in the FlashSystem 7300 IBM Documentation to find out what the error data means and what to do about it.

About this task

Complete the following steps if the Spectrum Virtualize software is not running on a node canister.

Procedure

  1. Connect a monitor to the VGA port and a keyboard to a USB port of the node canister. Consider any error messages on the monitor.
    For example, was it unable to find a good device from which to boot (such as device: /dev/sda [SAT], 1 offline uncorrectable sector)? If the system is booting to a UEFI Shell prompt, review the boot order configured on the canister.
  2. If no useful messages display on the monitor, complete the following steps.
    1. Power off the node canister by pulling it a little way out of the enclosure.
    2. Wait for 1 minute.
    3. Push the node canister back in to the enclosure and close the release latches.
      The node canister attempts to power on.
    4. If the power LED comes on green, then watch the VGA monitor for any useful messages.
    5. If the VGA monitor does not show any useful messages, try the next step.
  3. Attempt to access the UEFI setup utility on the VGA monitor by pulling out and pushing back in the node canister and holding down the ESC or Delete key on the keyboard.
    If the Setup Utility displays on the monitor, complete the following steps.
    1. If the node canister fault LED is flashing, access the Bmc self test log from the Server Mgmt tab to look for a cause.
    2. Access the System Event Log from the Server Mgmt tab.
      Events in this log might help to pinpoint the problem.
  4. If by using the setup utility you are unable to pinpoint a broken component, or if the setup utility does not start, complete the following steps.
    It is best to initially investigate a fault with the adapters and DIMMs.
    1. Power off the node canister by pulling the node canister out of the enclosure.
    2. Remove the node canister from the enclosure.
      Place it on a workbench where you can remove the cover.
    3. Remove the PCIe riser card in slot 1.
      See Removing and replacing a PCIe riser.
    4. Replace the cover, push the node canister back in to the enclosure, and close the release latches.
      The node canister attempts to power on.
    5. If the Spectrum Virtualize software now boots and the node canister fault LED comes on with the canister status LED blinking, then the adapter that you removed might be broken. Repeat the steps with a different adapter until you find the broken adapter.
  5. If the Spectrum Virtualize software does not load with all of the adapters that are removed, complete the following steps.
    1. Power off the node canister by pulling out the node canister from the enclosure.
    2. Remove the node canister from the enclosure.
      Place it on a workbench where you can remove the cover.
    3. Remove the DIMMs, but leave in one DIMM per microprocessor (CPU).
      For example, leave the DIMM in the A0 DIMM slot of each CPU. See Removing and replacing a memory module.
    4. Replace the cover, push the node canister back in to the enclosure, and close the release latches.
      The node canister attempts to power on.
    5. If the Spectrum Virtualize software boots and the node canister fault LED comes on with the canister status LED blinking, then one of the DIMMs that you removed might be broken. Repeat the steps with a different set of DIMMs in slot A0 until you find the broken DIMM.
  6. If you do not find any evidence of a broken DIMM or adapter, replace the node canister because a CPU or the system board might be broken.