![]() |
![]() |
||||
I also wanted to mention the latest JDK 1.6.0_u21 which has been installed now for the last 2 weeks. In some mathematical analysis software we use internally, we have noticed a large performance increase in this new release, probably due to the new HotSpot enhancements.
The software runs two separate calculations over and over again with different parameters, while analysing data. The average timings of these loops have dropped as follows: 104ms -> 86ms and 47ms -> 31ms.
We have also seen reductions in the Tomcat VM startup time of various VMs.
It would be interesting to see a benchmark of the various JDK minor releases, as this is a significant improvement.
Due to continual problems with Horde Webmail, we are pleased to announce that we are upgrading to two new webmail systems.
Please try out the new software and let us know what you think. We will retain Horde on http://horde.metawerx.net for customers who prefer the old interface, although we will no longer be upgrading the older software.
The following systems have been upgraded to the latest stable releases:
Apache have pre-released the long-awaited Tomcat 7, now in beta version 2, so we are pleased to announce we will also now support Tomcat 7, starting from this beta release.
To quote the Apache site, "This version of Apache Tomcat supports the Servlet 3.0, JSP 2.2 and EL 2.2 specifications. In addition, it includes numerous other improvements such as web application memory leak detection and prevention, extensive internal code clean-up and support for including external content directly in a web application (aliases)."
phpPGAdmin has been upgraded to version 4.2.3
Today I thought I'd write a news article describing a bit about some of the maintenance work we perform here at Metawerx...
A few weeks ago, we started the transition to the new Ubuntu release, and almost every server has had at least one problem so far. It's a new kernel, there are some new features, but I've read a lot on the net about display problems experienced by desktop and notebook users due to new graphics drivers and we've experienced a series of issues here as well. None of them have been serious enough to revert back to earlier revisions, but it's been a much slower upgrade process than previous releases of Ubuntu which have always been excellent.
To start with, a new display driver has affected the use of our DRAC cards. These cards let us get full keyboard/video/mouse into the servers remotely, which is necessary sometimes unless we want to live at the data center (which we don't). For example, if a server doesn't reboot correctly, we can bring up the monitor and gain full keyboard access to fix the problem either in the POST utilities or in a Linux recovery shell. With Lucid however, after Grub boots a kernel, the display changes into a graphical mode that isn't supported by the DRAC cards. We've solved this problem by forcing the framebuffer (fbcon) to be disabled, which sets the display into permanent text mode. That was also problematic, as the module disabler systems didn't seem to actually disable fbcon properly. However, we've worked out how to do it with a boot option and now have full access again.
Next, one server was locking up during the rc scripts. These scripts start all the processes and eventually the SSH service which we use to administer the boxes. The problem was a script that sets console fonts, after disabling it the boot now continues correctly (we have logged an Ubuntu bug about this issue but it doesn't seem to affect anyone else yet). This happened on our primary mail server, so the failover mail server was used temporarily during the upgrade and subsequent fix.
After that there was a problem with MySQL. Lucid forces the use of MySQL 5.1 by default. We could have used the Ubuntu backports repository to use MySQL 5.0, but want to keep inline with the current Ubuntu baseline. MySQL 5.1 is great, but one option in the my.cnf file (skip-bbd) was preventing MySQL 5.1 from starting. The first server upgraded after our test machines was used for MySQL replication, so there were also some minor replication warnings as the source server was still 5.0. After that, we had an issue with a JBoss cluster which required mixed or row-based replication. MySQL5.1 uses a new statement-based replication method by default so we needed to identify that issue and force it use mixed-mode otherwise it refused to write binary logs, and therefore refused to do very much at all.
Finally, today we performed a minor kernel upgrade from 2.6.31-20 to 2.6.31-22. On reboot the root partition refused to mount. Due to our high-security procedures which prevented the use of root as log-in account, and the inability to mount the root partition, we couldn't even get in via a Linux maintenance shell. This seems to be a new check in the new kernel which identifies that the root partition was at some time in data=ordered mode and we had it specified in fstab to mount as data=writeback mode so the new kernel was disallowing the mount entirely stating that data=writeback was a bad option for the filesystem. A tune2fs fixed the problem and now that server is booting properly on the new kernel.
Hopefully there will be no more problems. Earlier Ubuntu releases have been so smooth to upgrade to and Ubuntu's strict testing procedures are excellent, which is why Ubuntu has become the Linux OS of choice in recent years. This time has been a headache though, so many servers, so many issues and each of them very different, but as I said, nothing we haven't been to handle yet (fingers crossed!)
And.. Lucid Lynx does reboot very very quickly compared to earlier releases.
For any of our customers who are upgrading their dev boxes in the office or at home, if you have a severe problem with the upgrade, let us know as we may have run into it already.
Neale
The following systems have been upgraded to the latest stable releases:
We will also soon be migrating to MySQL 5.1 as a requirement of the upgrade to Ubuntu Lucid Lynx.
If you have any questions, please don't hesitate to contact support.
The following systems have been upgraded to the latest shiny new stable releases:
Tomcat 6.0.26 contains a new feature for checking web applications for memory leaks.
If you have any questions, please don't hesitate to contact support.
The following systems have been upgraded to the latest stable releases:
The Tomcat upgrades include important new security patches and increased performance.
If you have any questions, please don't hesitate to contact support.
We now have an emergency failover mail server in the US in case of power outages in Melbourne. This is kept up to date with copies of all mail from the primary mail server and can be brought online manually in case of an extended outage.
This is in addition to our fully redundant mail servers in Australia which use automatic failover via DRBD, Heartbeat, Tomcat and our proprietary mail routing systems to keep mail services running during hardware issues, upgrades or system restarts on the primary SMTP, POP3 and IMAP servers.