Adrian Dimcev's Blog

Persist the RX/TX NIC buffers settings across a reboot on Centos 7

Thu, 8 Mar 2018 09:39:00 -0700

You can easily verifiy/set the RX/TX NIC buffers using the ethtool utility (run yum install ethtool -y in case you don't have it installed):

- Verify the settings with -g (to also find the hardware maximums):

ethtool -g p1p1

- Set with -G to 4096 (in the above example they were already set at max by me previously, default they were 512):

ethtool -G p1p1 rx 4096 tx 4096

But we may want to make sure these settings persist across reboots. For this we can use the ETHTOOL_OPTS options in our NIC's settings.

First make sure that Network manager is disabled, run chkconfig --list NetworkManager

Then edit your interface configuration with your favorite editor:

vi /etc/sysconfig/network-scripts/ifcfg-p1p1

And add:

NM_CONTROLLED=no
(if you do not add this, the ETHTOOL_OPTS will be ignored, also make sure ONBOOT is set to yes).

ETHTOOL_OPTS="-G ${DEVICE} rx 4096 tx 4096"
(this line will set your buffers to desired setting, maximum in this case 4096.)

Save the changes made to the configuration file.

You can quickly check if your settings apply with:

ifdown p1p1 && ifup p1p1 and then ethtool -g p1p1

Ps. If you want to add multiple options separate them with ;:

ETHTOOL_OPTS="-G ${DEVICE} rx 4096; -G ${DEVICE} tx 4096"

Uninstalling NSX components on a cluster hung/stuck in Uninstalling

Wed, 7 Mar 2018 06:12:00 -0700

On a VMware cluster where the NSX components were installed on the hosts, we had them removed. The bits where removed from the hosts, the hosts were also rebooted, but the NSX UI in vSphere Web Client shows the Uninstalling to appear stuck: the Installation Status on the cluster says Uninstalling while on the hosts the status is In progress ... and spinning; and this never completes.

The solution provided by VMware support was:

- go in vSphere Web Client to Administration:

- There click vCenter Server Extensions, and then click vSphere ESX Agent Manager:

- Under vSphere ESX Agent Manager click the Manage tab and you will see the Agencies installed per your clusters. In our case the stuck one was showing as Uninstalled. Right-click it and click Delete Agency. A few tasks will be started on the hosts. Wait a few, and then refresh the Manage menu. During the operations its Status might turn to Warning, and when all tasks will finish it will disappear from the list.

- Going back to Networking and Security menu, Installation sub-menu, the cluster should look OK.

VMware vSphere Spectre Vulnerability - Patches available since November

Fri, 5 Jan 2018 12:19:00 -0700

With the info released about the Meltdown and Spectre vulnerabilities, in case you are running VMware vSphere, you may already have applied the patches mentioned in VMSA-2018-0002, maybe even beat them on AWS (especially if you are on vSphere 6.0).

For example the patch for ESXi 6.0, ESXi600-201711101-SG, was silently released in November last year, on November 09, 2017, as can be seen from Update Manager below or their KB 2151132.

vSphere 6.5 got the patch ESXi650-201712101-SG somewhat later, per their KB 2151099, on December 19, 2017. ESXi 5.5 got the incomplete patch ESXi550-201709101-SG also early on September 14, 2017, per their KB 2150876.

As a small note, for example ESXi650-201712101-SG, being a security release only, is included in the ESXi600-201711401-BG release as well which is a general bulletin (a general bulletin contains both the new security and bug fixes on top of the last version of the VIBs, see KB 2041654 for explanation). So installing ESXi600-201711401-BG in tandem with ESXi650-201712101-SG will make the SG appear as "obsoleted by host.