Dao's BLOG

This is related with a version compatibility issue.

nano +26 /usr/share/ganglia/cluster_view.php

change

$context_metrics = “”;

to

$context_metrics = array();

will do the job.

A zombie is already dead, so you cannot kill it. To clean up a zombie, it must be waited on by its parent, so killing the parent should work to eliminate the zombie.

An example of how you might send a signal to every process that is the parent of a zombie (note that this is extremely crude and might kill processes that you do not intend. I do not recommend using this sort of sledge hammer):

kill $(ps -A -ostat,ppid | awk ‘/[zZ]/ && !a[$2]++ {print $2}’)

Linux default greeters list user names on log in screen. This is sort of a security risk. Using lightdm + slick greeter can avoid this, rendering it back to the key-in username type.

To do so,

dnf install lightdm slick-greeter

systemctl stop gdm (or whatever dm you are using)

systemctl disable gdm

edit /etc/lightdm/lightdm,conf, change greeter-session to

greeter-session=slick-greeter

systemctl enable lighdm

if masked, systemctl unmask lightdm first

then reboot.

When you are working on a remote server through nomachine or vnc, and there is no video card on the remote server that supports OpenGL > 1.4, the MESA in default render OpenGL version 1.4. This will cause the application to fail even if they do not really need to raise a GUI.

Here is a way to force MESA to report its GL version to 3.3 such that certain applications will generate correct result without terminate while sensing a too low GL version:

Put the following environment string in front of your command:

MESA_GL_VERSION_OVERRIDE=3.3 your_command

This method does not work through ssh x-tunneling since in the latter case the x-server tends to take the GL command.

If you want this to go through an x-server that also uses mesa, you will need to put on the MESA_GL_VERSION_OVERRIDE=3.3 before your ssh command too.

Nomachine uses port 4000, however, the server may not have that port opened to your local IP. However, as long as you can ssh the server, you can use the following ssh tunnel to use Nomachine – securely and fast:

1. From a ssh client, do: ssh -L <port>:localhost:4000 -N -f <username>@<nomachine_server>
2. From your nomachine client, set host localhost, port of the port number as <port>

Here you go.

Scponly from CentOS 7 cannot chroot. Problem is that /sbin/scponlyc does not have corrent SUID. Do a

chmod 4755 /sbin/scponlyc

resolves it.

Ganglia web interface cannot display charts in CentOS7, if php is updated to version 7.

This is because /usr/share/ganglia/cluster_view.php in line 34, $context_metrics[] was originally initialized as “”, in line 26.

If changing line 26 to

$context_metrics = [];

The problem is fixed.

ssh to a linux box slow after keyed in password, taking 10+ seconds. If using ssh -v, found all the waiting spend on debug info:

debug1: pledge: exec

This usually happen when something stuck on the systemd-logind service.

Restart this service by running

systemctl restart systemd-logind

usually resolves this without rebooting the box.

This happens when systemd attemps to reload some previously saved state but was not able to. By “kill -1 1” to let it reload itself, systemctl works again without rebooting.