dennis henry

Install Yum on minimal CentOS

dennis — Wed, 06 Jul 2011 01:41:40 +0000

Hello all!

I ran into an interesting issue with a “minimal” CentOS installation and had to help a colleague install Yum on a server where only rpm existed. Here’s the final RPM line I used:

rpm -Uvh http://mirror.centos.org/centos/5/os/x86_64/CentOS/yum-3.2.22-37.el5.centos.noarch.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/python-elementtree-1.2.6-5.x86_64.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/python-iniparse-0.2.3-4.el5.noarch.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/python-sqlite-1.1.7-1.2.1.x86_64.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/rpm-python-4.4.2.3-22.el5.x86_64.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/python-urlgrabber-3.1.0-6.el5.noarch.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/yum-fastestmirror-1.1.16-16.el5.centos.noarch.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/yum-metadata-parser-1.1.2-3.el5.centos.x86_64.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/rpm-4.4.2.3-22.el5.x86_64.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/rpm-libs-4.4.2.3-22.el5.x86_64.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/popt-1.10.2.3-22.el5.x86_64.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/m2crypto-0.16-8.el5.x86_64.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/libxml2-2.6.26-2.1.12.x86_64.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/logrotate-3.7.4-12.x86_64.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/expat-1.95.8-8.3.el5_5.3.x86_64.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/nss-3.12.8-4.el5_6.x86_64.rpm http://mirror.centos.org/centos/5/os/x86_64/CentOS/nspr-4.8.6-1.el5_5.x86_64.rpm

This should install Yum and all the required depedancies for CentOS 5 x86_64 (at the time of this writing). Once this is ran, yum should be available to install any other software needed.

OpenSSH won’t start

dennis — Sat, 18 Jun 2011 17:19:56 +0000

I ran into an issue with OpenSSH Server would not start after fresh installation of it on a minimal CentOS 5 x86_64 VPS:

[root@vps ~]# /etc/init.d/sshd restart
Stopping sshd:                                             [FAILED]
Generating SSH1 RSA host key:                              [FAILED]

I delved a little further and discovered that ssh-keygen was failing:

[root@vps ~]# ssh-keygen
cannot read from /dev/urandom, No such file or directory

And finally, from there, I generated the urandom file:

/sbin/MAKEDEV urandom

Also, to ensure on restarts the urandom file (as well as pty and tty) are created, I added the following to /etc/rc.d/rc.local:

/sbin/MAKEDEV pty
/sbin/MAKEDEV tty
/sbin/MAKEDEV urandom

Hopefully this helps if you run into a similar issue down the line!

Allocate x IPs to OpenVZ VEs

dennis — Mon, 02 May 2011 15:52:57 +0000

I ran into an interesting issue today where we had a client migrating to us that needed us to allocated X IPs to each VPS on a node he had with us. X ranged from 1 IP to up to 12 per VE so I wrote a script that took three input files (one with available IPs [newips], one with VEIDs [veids], and one with the number of IPs each VEID was allocated [numips]) and doled out the IPs as needed. Here is the result:

#!/bin/bash
 
let r=1
let k=1
 
for i in `cat veids`
do
    numips=$(sed -n "$k"p numips)
        for (( c=1; c<=$numips; c++ ))
          do
                ip=$(sed -n "$r"p newips)
                echo "Add $ip to $i"
                vzctl set $i --ipadd $ip --save
                let r=r+1
        done
    let k=k+1
done

Hopefully this will help any others who run into a similar issue.

Have a great day!

Common Issues – No Networking on new OpenVZ VPS

dennis — Wed, 23 Mar 2011 22:42:21 +0000

Hello all,

Today I ran into an interesting issue where a VPS had no networking either inbound or outbound after setup. This is the error that would present when restarting network via init.d:

Bringing up interface venet0:
SIOCADDRT: Network is unreachable
SIOCADDRT: Network is unreachable

This issue also presented in that cPanel licenses could not activate not could any ping or host commands work properly. I eventually narrowed down the issue to the fact that the IPs that had been added to the VPS were not properly set to ARP by checking with the following command (run on the node):

arp | grep [INSERT IP HERE]

In order to resolve this, all I needed to do was add the arp entries for the IPs on the actual virtual environment by running the following commands on the node:

arp -s [INSERT IP HERE] `ifconfig eth0 | grep eth0 | awk '{print $5}'` pub

Make sure you run the command above for every IP on the VPS. This resolved the issue and all networking started working again properly.

Duplicate FTP Hosts in Show IP Usage

dennis — Wed, 08 Dec 2010 19:00:11 +0000

Ran into an issue today where a domain was showing in the cPanel “Show IP Address Usage” twice in the “FTP Usage” column. This was distressing the client so I looked into it and it appears this list is pulled from the /etc/pure-ftpd/ directory, which has the following directory structure:

root@server(/etc/pure-ftpd)$ ll
total 16
drwxr-xr-x  2 root root   4096 Dec  8 13:42 ./
drwxr-xr-x 90 root root  12288 Dec  8 13:42 ../
lrwxrwxrwx  1 root wheel    25 Dec  8 13:40 1.2.3.4 -> /home/user1/public_ftp/
lrwxrwxrwx  1 root wheel    24 Dec  8 13:40 2.3.4.5 -> /home/user2/public_ftp/
lrwxrwxrwx  1 root wheel    25 Dec  8 13:40 3.4.5.6 -> /home/user3/public_ftp/
lrwxrwxrwx  1 root wheel    25 Dec  8 13:40 4.5.6.7 -> /home/user4/public_ftp/
lrwxrwxrwx  1 root wheel    25 Dec  8 13:40 5.6.7.8 -> /home/user5/public_ftp/
lrwxrwxrwx  1 root wheel    23 Dec  8 13:40 6.7.8.9 -> /home/user3/public_ftp/

In my case, user3 had two IPs pointing to his public_ftp directory which caused the issue. I removed the symlink and then the IP Usage showed properly.

Just a quick fix for anyone else who experiences this issue!

Issue with GeSHI and CakePHP

dennis — Tue, 30 Nov 2010 17:12:20 +0000

So today I was having an issue where, for some reason, an incredibly long script in our script repository built using CakePHP would not display the GeSHI syntax-highlighted code. I tore apart the code looking for an answer and the only unique attribute to this script was its very long length (around 1k lines, over 50k characters). I eventually narrowed it down to the preg_replace that was being used in the CakePHP wrapper for GeSHI I was using and discovered that the pcre.backtrack_limit was too low at the default of 10000 (10 thousand). I upped the limit to 10000000 (10 million) and it’s now able to display the syntax highlighted code without issue.

Just thought I’d make a quick post because I could not find another blog with this solution. Hope this helps someone else in a similar predicament!

Total Memory on a Xen Node

dennis — Mon, 22 Nov 2010 17:53:36 +0000

I ran into an interesting issue the other day that I figured I’d make a post about in order for others. Since my company is getting more into Xen Hosting, I’ve tried to be at the forefront of the virtualization software for the company. We did run into an interesting issue though the other day when it came to determining the total amount of memory on a hardware node. Since we set up the Domain-0 to use only 512MB of memory, we were having a tough time determining how much total RAM is in the server. After googling for awhile and asking some co-workers, I discovered the following command:

xm info

This command shows quite a few interesting metrics including the total memory for the server. Using this I was able to easily discern the total amount of memory and write a quick script to compute total amount of memory:

#!/bin/bash
 
total_memory=$(xm info | grep total_memory | awk '{print $3}')
remaining=$(xm info | grep free_memory | awk '{print $3 - 512}')
used_memory=$((total_memory-remaining))
 
 
echo -e 'Total Memory:\t'$total_memory'MB ('$((total_memory/1024))'GB)'
echo -e 'Memory Used:\t'$used_memory'MB ('$((used_memory/1024))'GB)'
echo -e 'Remaining:\t'$remaining'MB ('$((remaining/1024))'GB)'

Hope this helps anyone else trying to figure out accurate Xen Memory usage

Install Xen 3.4.3 on CentOS 5.x

dennis — Fri, 10 Sep 2010 18:30:45 +0000

So I have had to install Xen 3.4.3 recently 3 different times so I thought I’d make a quick post regarding the best route I have found to do this. Please note, this may not be the most optimal route but it has worked well for me 3 times now:

yum install xen PyXML mesa-libGL
yum remove xen python-virtinst libvirt-python libvirt xen-libs
mkdir rpms
cd rpms
wget http://www.gitco.de/repo/xen3.4.3/libvirt-0.7.0-6.el5.x86_64.rpm
wget http://www.gitco.de/repo/xen3.4.3/libvirt-client-0.7.0-6.el5.x86_64.rpm
wget http://www.gitco.de/repo/xen3.4.3/libvirt-python-0.7.0-6.el5.x86_64.rpm
wget http://www.gitco.de/repo/xen3.4.3/python-virtinst-0.500.0-1.el5.noarch.rpm
wget http://www.gitco.de/repo/xen3.4.3/xen-3.4.3-1.el5.x86_64.rpm
wget http://www.gitco.de/repo/xen3.4.3/xen-libs-3.4.3-1.el5.x86_64.rpm
wget http://www.gitco.de/repo/xen3.4.3/glusterfs-client-2.0.8-1.el5.x86_64.rpm
wget http://www.gitco.de/repo/xen3.4.3/glusterfs-common-2.0.8-1.el5.x86_64.rpm
wget http://www.gitco.de/repo/xen3.4.3/fuse-2.7.4-1.rf.x86_64.rpm
wget http://www.gitco.de/repo/xen3.4.3/libibverbs-1.1-1.x86_64.rpm
rpm -ivh *.rpm

After that block, make sure you edit /etc/grub.conf to set the following lines:

---snip---
default = 0
---snip---
kernel /xen.gz-3.4.3 dom0_mem=512m
---snip---

The first will set the xen kernel as the default and the second will limit dom0 to only be able to use 512m of memory, leaving the rest for your VEs. After these changes, go ahead and reboot. Let me know if this route works for you!

exim: error while loading shared libraries: libperl.so: wrong ELF class: ELFCLASS32

dennis — Sat, 28 Aug 2010 20:22:07 +0000

So I ran into a very odd issue with a server today where exim would not start after a CentOS upgrade. The error message was such:

exim: error while loading shared libraries: libperl.so: wrong ELF class: ELFCLASS32

This error took awhile to google around for and I couldn’t find any relevant articles. The problem seemed to obviously be the libperl.so file but perl seemed fine and the file was identical to other server setups with the same version of perl, exim, cPanel, CentOS. I decided to try and rebuild exim with /scripts/eximup –force but that did nothing to help the situation. Eventually, I found the following command:

/scripts/eximup --source

To the best of my knowledge I gained from my research, this recompiles exim from source, converts to an rpm, then installs that. Sure enough, that fixed the issue and exim was running again with no issues. I figured I’d make this post due to the fact I couldn’t find one post with the information I needed.

Simple IPTables OpenVZ Setup

dennis — Mon, 02 Aug 2010 21:02:28 +0000

So after looking and failing to find a good article on how to describe the simple process to set up IPTables on an OpenVZ server, I figured I would write one here. The process is incredibly simple and can be broken down into 3 steps:

Empty out the contents of /etc/sysconfig/iptables
```
cat "" > /etc/sysconfig/iptables
```

Use the following line in /etc/sysconfig/iptables-config:

IPTABLES_MODULES="ipt_REJECT ipt_tos ipt_TOS ipt_LOG ip_conntrack ipt_limit ipt_multiport iptable_filter iptable_mangle ipt_TCPMSS ipt_tcpmss ipt_ttl ipt_length ipt_state iptable_nat ip_nat_ftp"

Use the following line in /etc/vz/vz.conf:

IPTABLES="ipt_REJECT ipt_tos ipt_limit ipt_multiport iptable_filter iptable_mangle ipt_TCPMSS ipt_tcpmss ipt_ttl ipt_length ipt_state"

After ensuring these three things, just stop both vz and iptables, start iptables, then start vz. You should then be able to use iptables within a virtualized container.