Thursday, December 15, 2011

Zimbra won't install on Debian testing

When attempting to install or upgrade Zimbra on Debian testing you will get a message regarding an incompatible platform (DEBIANUNKNOWN).

Change to your install directory (installdircetory/packages):

Repeat for each Zimbra Debian package

dpkg-deb -R package tempdir

In tempdir/DEBIAN/control change

Version: 7.1.3_GA_3346.DEBIAN5
Version: 7.1.3GA3346.DEBIANUNKNOWN

dpkg-deb -b tempdir package

(dpkg-deb will complain about the underscores, so remove them)

rm -R tempdir/*

After doing this for each package:

./ --platform-override


Tuesday, March 08, 2011

Koha z39.50 MARC record import fails

Koha is a fantastic open source product. Huge libraries run on it.

We suddenly (possibly after an upgrade) started getting first this error:

All arguments must be MARC::Field objects at /usr/share/koha/intranet/cgi-​bin/cataloguing/ line 151

And then suddenly this one:

Can't call method "add_subfields" on an undefined value at /usr/___/cataloguing/ line 142

This was resolved after days of hell by finding the solution to the second error here:

Resolve the error by changing the "z3950NormalizeAuthor z3950AuthorAuthFields" setting under the Cataloguing Tab, and Record Structure section to "don't copy".


Thursday, January 27, 2011

Wordpress database error

This one did my head in for a while.

After a Wordpress reinstall I could not get the new installtion to connect to a database on a remote server. The error message from the "wp-config.php" file was:

Error Establishing a Database Connection

Everything in the file seemed correct. Wordpress was unable to write the file and had asked me to paste the configuration into the wp-config.php file after creating it.

So I eventually deleted this and modified a copy of the "wp-config-sample.php" file.

Presto! It worked. I believe the issue may have been the encoding of the copy and pasted configuration.

Labels: , ,

Saturday, December 11, 2010

Zimbra install on Debian Testing

Nightmare install with Zimbra on Debian Testing.

I kept on encountering the following error:

Creating server entry for {myserver}...failed

and in the detailed log:

ERROR: service.FAILURE (system failure: ZimbraLdapContext) (cause: javax.naming.CommunicationException {myserver}:389)

The standard diagnosis is a DNS / host miconfiguration:

I tried everything I could relating to fixing a DNS and host configuration without any luck.

I then found this:

The problem is a resolution error because Java is trying to use IPv6.

To disable IPv6, from

echo net.ipv6.conf.all.disable_ipv6=1 > /etc/sysctl.d/disableipv6.conf

Presto. Works.



Wednesday, February 03, 2010

IRQ / interrupt conflict issues with megaraid and Linux

Ever since Linux kernels beyond, I've been unable to boot my HP Netserver LH3.

It gives a kernel panic with a variety of messages that always include "common_interrupt+0x38/0x39"

I tried everything from updating the BIOS and firmware, to myriads of differing kernel configurations (at 4 hours per compile, imagine the agony).

I also tried booting with every combination of the kernel boot parameters acpi=off, noapic, nolapic, irqpoll, etc that I could imagine. No dice.

In final desperation, I began unplugging PCI cards one by one before finally disconnecting the SDLT drive from the Netraid SCSI. Et voila!

It seems that in later kernels or versions of udev, sharing the SCSI bus with the LH3 drives causes conflicts.

I have an additional AHA SCSI card in the machine, so will try connecting the SDLT to that and see how it goes.


Tuesday, February 02, 2010

Applying firmware update floppies

One of our servers is an old HP Netserver LH3. It still works well, but gave some issues with a recent linux kernel update. So in an effort to revive it, I sought out the last firmware updates.

HP Netserver LH3 firmware updates

The problem with these is that they download as self extracting stiffy disk images. Hands up if you can still write to one of those...

The first step was to download Winzip and extract the files. The next was far more complicated. To access the files on the actual image (.img file), I eventually found this neat win32 application VFD that works under Vista. It allows you to mount the IMG image as a drive, from where you can access the files (you must run the winvfd file as administrator) from a administrator command prompt.

Lifesaver! From there I copied the files to a bootable CD (I got the a Windows 98 DOS CD image from All Boot Disks) and ran them on the machine.

A small headache was that one of the firmware updates was incompatible with HIMEM.SYS. Luckily I found an old Windows 95 startup disk that does not load himem and used that to boot. It recognised the CD drive allowing me access to the firmware updates I'd cut onto CD.


Sunday, January 24, 2010

Debian Testing update nightmare

I applied updates to Debian testing for the first time in a while. A whopping 1,5GB set of package downloads.

I'm at fault for a few things here:

  • I have not been regularly applying updates - last one was probably in October.

  • I should not be running a production server on the testing release - I do this to access the latest R libraries amongst others

  • I should not have done this the weekend before i needed to be at an offsite client workshop for a week

Those admissions made, what a nightmare!

The Grub2 completely broke the system. The source of the problem is that Grub2 appears to have real problems on RAIDs. I got the "Cannot find operating system" error. Using the SysRescCD I managed to rewrite the MBR, uninstall Grub2, revert to Grub Legacy, and finally bring the system up again.

The next disaster? X.Org locks up entirely after the boot. You cannot use the keyboard or mouse and you cannot switch to another TTY. SysRescCD again, and disabled GDM/KDM in the /etc/init.d startup scripts.

Biggest disaster? The server is running on an old HP Netserver LH3 with a NetRaid controller. I've been using the kernel for a while now because later kernels' Megaraid drivers have given problems with the card causing kernel panics. The problem? Udev now gives the message: "udev requires a kernel >= 2.6.26"

I'm busy compiling the kernel from scratch at the moment with Megaraid legacy support. Let's hope I can get it up. t-2 hours from departure to the client offsite with a koi pond crisis to sort out too.



I shmaak SA Blogs, sorted with