Known issues and limitations in Bare Metal Solution

This page describes known issues and limitations that you might run into when using Bare Metal Solution.

Limitations

Modification to BIOS settings is not supported

Bare Metal Solution does not support modifications to BIOS settings, including disabling C-states and P-states at the BIOS level.

To work around this issue, you can utilize the OS-level controls of C-states and P-states through your OS's power management functionality. To learn how to do this, refer to your OS documentation.

VPC Service Controls is not supported

Bare Metal Solution does not support VPC Service Controls. Connecting a VPC with service controls enabled to your Bare Metal Solution environment does not uphold any service control guarantees.

The Bare Metal Solution API can be added to a secure perimeter. However, the VPC Service Controls perimeters do not extend to the Bare Metal Solution environment in the regional extensions. For more information, see Supported products and limitations.

If you still want to use Bare Metal Solution with VPC Service Controls enabled, contact Customer Care to add your Google Cloud project to allowlist for using this feature.

Changing ASN is not supported

Bare Metal Solution and Partner Interconnect do not support custom autonomous system numbers (ASN). The Bare Metal Solution client network ASN is set to 65500. For Partner Interconnect, all Cloud Routers must have a local ASN of 16550.

We recommend that you plan your deployments accordingly.

Known issues

Cold shutdown after the first reboot through OVM Manager

For servers provisioned with OVM 3.4.6, the first reboot initiated through the OVM Manager, either after the initial provisioning or re-imaging, results in a cold shutdown.

To learn how to power on a server, see Operate your Bare Metal Solution server.

Server takes a long time to boot

The boot time may vary depending on the size of your server and the number of shared LUNs.

The larger the server, the longer it takes to boot.

The number of shared LUNs on a server also impacts its boot time. For instance, a o2-highmem-224-metal server with ~100 shared LUNs might take over an hour to boot compared to a o2-highmem-224-metal server with fewer LUNs that would take ~45 minutes. This is because of the time required to run all the checks and is normal.

Buffer overflow when using ethtool with the debug flag

A bug causing a buffer overflow when using ethtool -d was fixed in Linux 5.8 kernel. This bug can cause a kernel panic and might impact your Bare Metal Solution server depending on the OS and hypervisor that you are using.

Following are our recommendations and workaround for this bug:

Red Hat Enterprise Linux (RHEL)

If you are using RHEL 7.x, then follow these guidelines:

  • Do not run ethtool with the debug flag: ethtool -d.

  • Prevent the sosreport utility from invoking ethtool -d by disabling the networking plugin.

    In the /etc/sos/sos.conf file, add the following lines:

    [plugins]
disable = networking

If you are using RHEL 8.x, then follow these guidelines:

  • Update to kernel version RHEL 8.3 (kernel 4.18.0-240) or later.

  • For systems with kernel versions lower than 4.18.0-240 that can't be updated, prevent the sosreport utility from invoking ethtool -d by disabling the networking plugin.

    In the /etc/sos/sos.conf file, add the following lines:

    [report]
skip-plugins = networking

For more information, see the Red Hat Solution.

SUSE Linux Enterprise Server(SLES)

This bug was fixed in SLES 15 SP4 (kernel version 5.14.21-150400.22.1). Update to SLES 15 SP4 (kernel version 5.14.21-150400.22.1) or later.

Oracle Enterprise Linux

If you are using Oracle Linux 7.x, then follow these guidelines:

  • Do not run ethtool with the debug flag: ethtool -d.

  • Prevent the sosreport utility from invoking ethtool -d by disabling the networking plugin.

    In the /etc/sos/sos.conf file, add the following lines:

    [plugins]
disable = networking

If you are using Oracle 8.x, then follow these guidelines:

  • This bug was fixed in OL8U7 (kernel-uek-5.15.0-3.60.5.1). You can update to the most recent kernel by following the update instructions from Oracle.

  • If you can't update to kernel version kernel-uek-5.15.0-3.60.5.1, prevent the sosreport utility from invoking ethtool -d by disabling the networking plugin.

    In the /etc/sos/sos.conf file, add the following lines:

    [report]
skip-plugins = networking

Oracle VM Server

  • Do not run ethtool in dom0 domain with the debug flag: ethtool -d.

  • Prevent the sosreport utility from invoking ethtool -d by disabling the networking plugin.

    In the /etc/sos/sos.conf file, add the following lines:

    [plugins]
disable = networking

OVM server can't connect to OVM Manager after TS54 firmware upgrade

If you have upgraded your Bare Metal Solution server running Oracle VM (OVM) and it cannot connect to the OVM Manager, then this issue might be the reason. Upgrading firmware of a Bare Metal Solution server changes its SMBIOS UUID. The OVM uses this UUID to identify itself with the OVM Manager. Therefore, the change of UUID can cause issues with the communication between the two. To prevent this, before upgrading the firmware of your Bare Metal Solution server, apply the workaround described in Oracle Doc ID 1534416.1. For assistance in implementing this workaround, contact Customer Care.