Arrfab's blog Some tips and tricks, mostly around CentOS

This is a question we receive a lot on the CentOS forum : "i've updated my system with yum and it installed a new kernel but it didn't start on my default kernel ..." Ok, so let explain what happened for sure : you've installed initially a box with multiple kernels, for example kernel and kernel-xen (it can be PAE as well ...).
Of course you know how to configure grub and you modify the /boot/grub/grub.conf to put your default kernel to be the standard one (because you installed the kernel-xen for testing purposes only ....)

Everything is fine ... until next kernel update : it reverts the default kernel to boot in grub to the kernel-xen ...

Explanations : When you installed both kernels, a file containing a default kernel for the system has been created (/etc/sysconfig/kernel .. check for defaultkernel=) .. and if you installed kernel-xen (for a test maybe ...) this is the default kernel.  After a new kernel installation, /sbin/new-kernel-pkg is called to make a new initrd and use grubby to automatically add the newer kernel entry in the grub config file ... If you have a look in this new-kernel-pkg script, you'll see that he uses the /etc/sysconfig/kernel file to …

Thanks to sdog's blog , i've found and tested Zattoo . They provide binaries (yep, binaries only ...) for Mac OS X, also for an OS from Redmond, and ... for Linux ! You have the choice between a .deb package, a RPM package , and a .tar.gz file containing the binaries. Of course i tested the rpm, that was packaged very badly .. `rpm -qi zattoo` gave me quickly the answer why : '(Converted from a deb package by alien version 8.64.)' :o)

I'm wondering if i need/can (due to their proprietary license) re-package the static binaries with the correct 'Requires:' field and a '%post' section doing all the symlinks (yep, their binaries are build against/for .so.0d :-( ....) and a proper ldconfig ... and put a .nosrc in the package name ....

Anyway, the future will tell if the Zattoo service will remain free or not ...

I've just joined the mugshot CentOS group so i needed the mugshot client on my gnome desktop ... here it is for the i386 platform and here for the x86_64 one

I'm still wondering what will be the benefit of mugshot, but time will tell ... :-)

I work for a company that uses IBM iSeries machines (aka as400) so i needed a tn5250 emulator. CentOS 5 comes with tn5250 but that package is very limited. I found tn5250j (java based) and i've tested it for several months ... now it's time to package it as a rpm instead of using their java jar installer ... Here it is : http://rpms.arrfab.net/centos/5/i386/repodata/repoview/tn5250j-0-0.6.0-1.el5.af.html  .

You need the sun JRE 1.4 minimum (i use the 1.5 without any problems). This package doesn't work correctly with the provided  java-1.4.2-gcj-compat !

I had to setup a CentOS box that needed access to multiple VLANs configured in a bunch of Cisco switches ... Instead of having a different physical adapter for each vlan in linux, you can configure your physical interface with multiple logicals interfaces that will be 'tagged' (aka 802.1q vlan tagging). You need of course to configure your switch so that your eth0 will be in trunk mode ('switchport mode trunk' on cisco) and then configure logical interfaces on the linux side. A quick `grep -Ri vlan /usr/share/doc/initscripts-8.45.14.EL/` showed me that on CentOS you just need to create a /etc/sysconfig/network-scripts/ifcfg-eth0.30 (assuming that the vlan you need access to is vlan 30) and this file will have to look like this :

DEVICE=eth0.30
BOOTPROTO=STATIC
IPADDR=10.111.32.23
NETMASK=255.255.255.240
VLAN=yes
ONBOOT=yes

Of course you can fix parameters (dhcp, onboot, etc ...) like usual. A `ifup eth0.30` will bring this logical interface up (and will automatically modprobe the 8021q module) and you'll have access to machines sitting in this vlan 30 from your single eth0 physical inteface. Of course you can create multiple …

I've recently installed a IBM Domino 7 cluster on CentOS 4.5. I used Heartbeat and DRBD for data replication between the two nodes. I first started with a v1 heartbeat cluster style (using haresources) but then i decided to swith to v2 (using crm / cib.xml). The problem is that the init script for domino that i found somewhere on the ibm website didn't answer to the 'service domino status' and was not even returning code ... so when you wanted to move a resource from one node to the other, the domino server was always listed as an unmanaged resources and failed to switch to the other node. So i attach the modified domino init script.

I had also to put a timeout value for the stop operation in the cib.xml. Otherwise heartbeat complained that he had to kill the process itself (domino is slow to stop .... :o)). You can do this by entering the following parameters for the lsb resource :

operations op id="afe9e8fe-bc40-4b3f-b563-9bd70e695ab6" name="stop" timeout="120s" start_delay="0" disabled="false " role="Started" /operations

Hope that this will help someone ... i'll try to write a full doc (centos/heartbeat/drbd) on the wiki very soon ....

IBM released Notes client version 8 recently. Version 7 for Linux sucked but this newer version is better integrated, verify dependencies and i'm using it now for one week without any problems so far ... Still not delivered in rpm format (when will IBM/Lotus understand the benefit of RPM instead of their stupid Installshield anywhere java app ... ?) though. On the other hand, it contains the IM Sametime plugin (very well integrated this time) so i don't need to use Gaim/Pidgin meanwhile plugin to connect to our Sametime server . IBM decided to package OpenOffice (ibm modified version) inside the Notes client so it seems their philosophy is to push Odt format ...

I needed a tool to synchronize imap folders from one server (M\$ Exchange) to another imap server. Imapsync seemed to be a good tool but didn't exist for CentOS 5 so i've built it : http://rpms.arrfab.net/centos/5/i386/repodata/repoview/imapsync-0-1.219-1.html   . Also available in the x86_64 tree ...

One of my customers wanted to setup Vmware server on top of CentOS 4.5 on a IBM HS20 (in an IBM BladeCenter). The guest machines were unable to connect to the network (or just briefly, like 3 or 4 icmp packets) but were able to at least see remote machines mac addresses (as shown through arp). On the other hand, everything was running smoothly in NAT mode . I had a look on upstream Cisco Switches configs but everything was ok ... So i decided to update the firmware of the built-in Broadcom Fiber ethernet cards and after that everything went back to normal in bridge mode. So it seems that these devices didn't support correctly the bridging mode out-of-the-box and i assume that it should have been the same with other virt products (qemu, kvm, xen, etc ...). It reminds me a colleague some years ago : when you told him that you had a problem, his answer was always : 'Have you updated the BIOS ?' simple but it works often ... ;-)

Yep, you've correctly read ... Oracle released a modified version of YaST to run on top of Enterprise Linux (including Unbreakable linux of course). Is it april 1st ? or another Oracle inconsistency ? Will their next unbreakable linux be based on SLES ? :o)