Posts Tagged ‘ xen

Silly things I decide to do…

I’m not sure what gets into my head at times but often I find something silly to work on to pass the time and this time around it was getting FreeBSD 9.0 installed running under Xen with full para-virtualization (PV). In the past (I think it was 8.0) I couldn’t even install FreeBSD as a HVM domU as there was some issue with the bootloader. I started seeing posts around the FreeBSD 8.2 release time frame with people getting FreeBSD running under a Linux Xen machine as domU both HVM and PV so I suppose this finally came up on my list of something to do? Anyway I sort of resolved this installation fix around FreeBSD 8.0 by creating a file based image on an existing FreeBSD system and compiling a FreeBSD XEN enabled kernel. Details for what I consider the old way can be found at the following location: HOWTO: Create a FreeBSD 8 i386 Xen PV domU

The new way I used today involves first installing FreeBSD as a HVM domU and transforming the domU from HVM to PV. The instructions were written for FreeBSD 8.2 and Xen 3.4 but I can confirm that they do work for FreeBSD 9.0 with a few small changes (mainly in block device name – it listed ad1s1a whereas I found I had to use ad1p2 as the main device) and using Xen 4.x. I’m going to read up on what’s new in 9.0 so perhaps devices are being renamed. The new instructions can be found at the following site: Xen FreeBSD 8.2 DomU (PV) — Step by Step Howto

Everything seems to be working and it’s nice to have something new to play around with, not sure how effective the installation will be as far as messing with ZFS goes. One thing I’m pretty sure I’ll run into issues with ZFS is the odd memory cap (850MB) the domU has. Last I heard things like dedup fail in low memory setups.

Until next time…

Highly Productive Weekend – meh?

It’s the designated Quarterly Maintenance weekend for my day job though not all that much I had to assist with since I’m only 50% on the sysdba team. So since I was sticking around for the weekend and there really isn’t all that much to do otherwise while my wife worked I took the time to get Oracle RAC setup in my home server environment again. My previous install was blown away quite a while back to make room for other cool projects.

This time around the entire install was based upon the latest release ( of the Grid Infrastructure and Database. Also this time around I setup the Xen guests as paravirtualized (pv) rather than full (hvm) and used CentOS 5.7 rather than Oracle Enterprise Linux. I hadn’t had experience setting up OEL as a pv but last Fall found a slick way to install CentOS. I’ll miss out on things such as ACFS but I have no desire to use ASM as a LVM (probably since the shared disk I use for ASM is essentially LVs over iSCSI – you can only virtualize so much before it’s pointless -no?)

Overall the entire setup was quite easy. Using open-iSCSI on the guests was the only real thing different from my 11.1 RAC installs. It seems like each RAC install since 11.2 has really been improved and has become quite flawless. I have no real use case for RAC but at least I have a recent setup around again to play around with.

I also decided to attempt my first installation of Oracle Enterprise Manager 12c. I had one other database server with an adequate amount of memory to attempt the installation but so far it’s been slow going (stuck at 66% for quite some time). I’m sure my ‘ignoring’ the 3GB physical RAM requirement will cause an issue since WebLogic loves to crash on low-memory machines (java… meh) At least I tried and suppose I could always shutdown the two RAC nodes to reclaim 4GB of physical RAM to ensure a successful installation.

So, there you have it, my weekend in less than 500 words.

Note: Summary of installation steps to be provided on Samdog Consulting site in the near future.

Update: Oracle Enterprise Manager 12c install did actually die upon first attempt. Successfully installed after server was brought up with 5GB physical RAM. It’s a pig but to be expected I suppose...

Long Weekend with Tech

With having a long(er) weekend I’ve been able to finally get to some small techy tasks done on the home server. The majority were just making sure all updates had been applied and others involved just moving things around to free up disk space.

One thing that impressed me was how pvmove works so well. I have a high level understanding of LVM but it’s impressive how you can just move things around in LVM without destroying data. This time around I did something a bit crazier (perhaps stupid but that’s what backups are for). I used pvmove to move data off a disk I wanted to retire while several resources were still actively using the volume group (i.e databases, domU hosts, etc…). It went slow, pvmove isn’t known for speed anyway, but it successfully worked. I would have thought a certain database would have freaked out but nope, not a peep from it.

I went a bit nuts with messing around with pygrub on some new domU machines. I hadn’t explored too much of Xen 4 since upgrading to Debian 6 a few months ago but am pleased that pygrub support is now part of xen-tools 4.2 (without having to use a pygrub role script). The only issue was with using xen-tools’s image create with pygrub is on a newer Ubuntu releases there is an issue after image creation with booting off xvda devices. For Ubuntu xen-tools installs the generic server kernel and upon boot it fails to mount a xvda devices (xvda is the default device xen-tools uses when creating images). Easy enough to fix, just change the 80-install-kernel script and change linux_kernel_package to “linux-virtual” rather than linux-image-${linux_kernel_type}. I noticed that under /usr/lib/xen-tools that symlinks existed for most newer Ubuntu dists which pointed to the karmic.d directory, I simply copied the karmic.d directory and made the modification to 80-install-kernel then changed symlinks to point to the new directory. I plan on checking out the newer xen-tools to see if this has been fixed as we all know that Debian tends to get a bit behind in versions. If it’s not there it’s worthwhile to report the issue to save others time. Also there seems to be an issue when specifying fs=btrfs in which it fails to create a btrfs filesystem. It appears to be something to do with the mkfs.btrfs options passed when creating the filesystem.

Finally, I have been getting into some podcasts as of recently. My recent favorities are:

 I’m sure there are others but at the moment I plan on getting caught up with recent podcasts from the two and see if I can keep up with them weekly. Thankfully Google Listen works well enough and can download steams in advance as it probably keeps AT&T from bugging me about data usage on my phone. dom0 upgrade complete!

Just a followup on the previous post. Loki (dom0) has been upgraded to Debian 6.x successfully. It was actually quite easy and as expected the domUs started right up after configuring Xen for bridging.

Now that there is a newer kernel I can experiment with things such as btrfs, etc… There should also be so speed ups as well 🙂

Happy Spring!

Xen: Debian Lenny Dom0 with Ubuntu Lucid Guest

This actually works and now I’m rockin 10.04