Removing and replacing a memory module (DIMM): 2145-SV2 and 2145-SA2

You can remove and replace a faulty dual in-line memory module (DIMM) from a node canister. You can also use this procedure to add DIMMs to increase the memory capacity of the node canister.

About this task

Notes:
  • Electrostatic discharge precaution symbol The memory modules are sensitive to electrostatic-discharge (ESD). Take precautions to avoid damage from static electricity. When you remove or replace this part, you must follow recommended procedures for handling ESD-sensitive devices.
  • No tools are required to complete this task. Do not remove or loosen any screws when you complete these procedures.
As Figure 1 shows, each node canister contains two processors, which are identified as CPU 1 and CPU 2.
Figure 1. Location of CPUs and DIMM slots
Location of CPUs and DIMM slots
  • A CPU processor has six memory channels, which are labeled A-F.
  • Each memory channel has 2 DIMM slots, numbered 0-1. For example, DIMM slots A0 and A1 are in memory channel A.
  • On the system board, the DIMM slots are labeled according to their memory channel and slot. They are associated with the CPU nearest to their DIMM slots.
  • To maintain consistent airflow and cooling, each DIMM slot must contain either a memory module or a filler blank.
  • You do not have to replace all of the memory modules at the same time. However, when installing memory modules, be sure that you follow the configurations that are described in Table 1.
    Table 1. DIMM locations and memory configurations
    DIMM slot Slot index Configuration of DIMMs and blanks
      128 GB 384 GB 768 GB
    F0 11   32 GB 32 GB
    F1 12     32 GB
    E0 9   32 GB 32 GB
    E1 10     32 GB
    D0 7 32 GB 32 GB 32 GB
    D1 8     32 GB
    CPU1  
    A1 2     32 GB
    A0 1 32 GB 32 GB 32 GB
    B1 4     32 GB
    B0 3   32 GB 32 GB
    C1 4     32 GB
    C0 5   32 GB 32 GB
     
    C0 17   32 GB 32 GB
    C1 18     32 GB
    B0 15   32 GB 32 GB
    B1 16     32 GB
    A0 13 32 GB 32 GB 32 GB
    A1 14     32 GB
    CPU2  
    D1 20     32 GB
    D0 19 32 GB 32 GB 32 GB
    E1 22     32 GB
    E0 21   32 GB 32 GB
    F1 24     32 GB
    F0 23   32 GB 32 GB
If a node error 510 is logged (error code 1022 in the system log), the sense data includes the following information, which you can use in addition to Table 1 to determine which DIMMs need to be replaced.
  • Detected memory size in MB.
  • Expected memory size in MB.
  • A list with one entry per DIMM slot, show the size of the DIMM in GB for each slot.

Procedure

Identifying failed DIMMs

  1. If you are replacing failed DIMMs, use the service assistant or the System Event Log in the management GUI to obtain the sense data that shows the detected and expected memory sizes and the size of the DIMM in each slot.
  2. Identify the column in Table 1 that contains the expected configuration by matching the expected memory size from the error sense data to the column that has the same node canister total memory.
  3. Compare the size of the DIMM reported in each slot in the error data with the expected DIMM size for the slot index and node canister total memory size that is shown in Table 1. For each mismatch, the DIMM must be replaced with the size of DIMM that is indicated in the table. See the replaceable units topics in the Reference section for the correct FRU part number for each size of DIMM.

Removing the node canister

  1. Remove the node canister, as described in Removing and replacing the node canister in the node: 2145-SV2 and 2145-SA2, and place it on a flat, level surface.

Removing the DIMM

  1. Remove the cover, as described in Removing and replacing the cover of the canister: 2145-SV2 and 2145-SA2.
  2. Locate the appropriate DIMM slot.
    As Figure 2 shows, the DIMM locations are identified on the system board.
    Figure 2. Identifying DIMM locations
    Example of DIMM locations printed on the system board
  3. Press the locking tabs on the side of the DIMM to eject it.
  4. Lift the DIMM up and out of the slot.
  5. If you have other tasks to complete inside the node, do those tasks now.

Replacing or adding the DIMM

Important: If you are adding memory to a node canister, you must remove that node from the system configuration before you start the following procedure. To do so, you can use the management GUI or the CLI.
  • To use the management GUI, right-click the node canister and select Remove.
  • To use the CLI, enter the following command, where object_id | object_name identifies the node canister that receives the additional memory:
    rmnodecanister object_id | object_name 

If you are replacing a faulty DIMM with a new one from FRU stock, you do not need to remove the node canister from the system configuration.

  1. Touch the static-protective package that contains the new DIMM to any unpainted metal surface on the node canister. Remove the DIMM from the package.
  2. Turn the DIMM so that the DIMM keys align correctly with the connector on the main board.
  3. Insert the DIMM into the connector by aligning the edges of the DIMM with the slots at the ends of the DIMM connector, as shown in Figure 3.
    Figure 3. Installing the DIMM
    Installing the DIMM
    Attention: If there is a gap between the DIMM and the retaining clips, the DIMM is not correctly inserted; open the retaining clips, remove the DIMM, and then reinsert it.
  4. Firmly press the DIMM straight down into the connector by applying pressure on both ends of the DIMM simultaneously. The retaining clips snap into the locked position when the DIMM is firmly seated in the connector.
  5. Repeat steps 10 through 13 until all the new or replacement DIMMs are installed.
  6. Install the cover, as described in Removing and replacing the cover of the canister: 2145-SV2 and 2145-SA2.
  7. Reinsert the node canister, as described in Removing and replacing the node canister in the node: 2145-SV2 and 2145-SA2.
    If the canister does not begin to power up, check that the canister is fully inserted into the enclosure.
  8. When the node is active again, go to the management GUI and look for any unfixed events that are related to the DIMMs.