Martello

We have an old rack server that runs a version of Solaris 10 for X86. For the most part, this machine never gives any trouble and we very rarely need to reboot. However, in recent months when we have rebooted it,  it has sometimes failed to boot, failing with a system console error that looks something like:

cannot read biosint
trap type 13 (0xd) err code 0x98b8 eip=0x0
...lots of register and stack trace data...
panic: corrupted boot archive . . . boot loader
Press any key to reboot

We found the instructions detailed here resolved this issue for us (although the error message is not exactly the same).

I was trialling a MySQL database upgrade today (from 5.0 to 5.1) and hit a nasty little gotcha that I feel warrants a mention (mostly to myself). In its simplest form, the upgrade process is very straightforward and involves simply loading data dumped from MySQL 5.0 into a 5.1 database (using a regular mysql command), followed by a once-off execution of the mysql_upgrade command (to bring the various tables and grants up to date):

$ mysql -u root -p testdb < dumped50.sql
$ mysql_upgrade -u root -p testdb

However, when I tried the latter command, I received this error:

Looking for 'mysql' in: mysql
FATAL ERROR: Can't find 'mysql'


I did fine a simple solution here which was to first change to the directory containing the mysql_upgrade program itself and execute it from there:

$ cd /opt/coolstack/mysql/bin
$ ./mysql_upgrade -u root -p testdb

As I said, its a simple solution and I do tend to follow this general rule most of the time but nevertheless, the error generated when you don’t do this is far for informative.

I needed to rename a zone on a Solaris 10 system earlier this week and here are some notes on how I did it.

The process of renaming a zone is essentially a task of renaming, editing and replacing strings in a series of (mostly XML) configuration files. All of the tasks below were carried out from the global zone on the system in question.

1. Shut down the zone to be renamed

# zoneadm -z <oldname> halt

2. Modify the configuration files that store the relevant zone configuration

# vi /etc/zones/index
Change all references of <oldname> to <newname> as appropriate
# cd /etc/zones
# mv <oldname>.xml <newname>.xml
# vi <newname>.xml
Change all references of <oldname> to <newname> as appropriate

3. Rename the main zone path for the zone

# cd /export/zones
# mv <oldname> <newname>

Your zone path may be different than the one shown above

4. Modify (network) configuration files of new zone

Depending on the applications installed in your zone, there may be several files you need to update. The essential networking files are:

# cd /export/zones/<newname>/root
# vi etc/hosts
# vi etc/nodename


But others containing your old host/zone name can also be found using this command:

# cd /export/zones/<newname>/root/etc
# find . -type f | xargs grep <oldname>

5. Boot the new zone again
# zoneadm -z <newname> boot

The Solaris Coolstack team have announced some important changes that impact on the future of the popular open source software stack. In a nutshell, they are combining their Cool Stack software (intended for use with Solaris 10) with their Web Stack software (intended for use for OpenSolaris) into a single software bundle, which will be known as Sun Web Stack.

Not only will this new stack be supported on Solaris 10 and OpenSolaris, but Linux support is also being added, as is full support for those using the software in production a capacity.

The first version of Sun Web Stack is scheduled for release in November 2008 and will follow the version numbering from Coolstack. Thus, the first release will be Sun Web Stack 1.4.

Source: Shanti’s Blog

I recently installed Squid, the HTTP Proxy, for the first time. The version I installed was 2.6-STABLE16 and came as part of Solaris Coolstack 1.2. Here are some notes from the installation process:

Installation

The application on Solaris x86 using the pkgadd utility and the appropriate Coolstack Squid package file (CSKsquid_1.2_x86.pkg). The installation process also created a new SMF service called csk-squid.

Setup

This was the first time I had installed Squid so, after reading the documentation provided by Coolstack, I tried starting the Squid service. Unfortunately, it kept complaining about missing cache directories and file permissions, and would not start. In the end, a colleague directed me towards the squid -z command, which carried out all of the required post-installation setup and duly sorted out all of the directory/permissions problems.

Configuration

I only had to make two small changes to the Squid configuration file (/opt/coolstack/squid/etc/squid.conf):

1. Squid uses port 3128 by default but, since our corporate firewall permits very few outbound ports, we modified squid to use port 8080 instead:

http_port 8080

2. We wanted to restrict access to the new HTTP Proxy to only a certain range of corporate IP addresses. This was easily achieved by adding a simple access rule as follows:

acl my_networks src XXX.XXX.XXX.XXX/24
http_access allow my_networks

Other than this, the process was surprisingly straightforward.