Volume 27, Number 1 - January 2011
|CACTUS Officer Elections
Thursday, January 20, 7:00 PM
The CACTUS Newsletter is a monthly publication, distributed to our members and other interested people. Visit the CACTUS Newsletter on the web at http://www.cactus.org/CACTUS/Newsletter/. There you will find archives of back issues, as well as instructions on how to subscribe to the e-mail distribution. We welcome newsletter submissions by our members. Please contact newsletter [at] cactus <dot> org for more information.
The next CACTUS meeting will be held on Thursday, January 20, 2011 at 7:00 PM (doors open at 6:30 PM for pizza and informal discussion) at the Mangia Pizza on Guadalupe. The address is 3016 Guadalupe Suite 100 (see http://tinyurl.com/ycsey5y). This is just north of 30th and Guadalupe, near the Wheatsville COOP. There is parking in the garage on the back.
First item on our agenda for Thursdays' meeting is Officer Elections. It is that time again... There are several reasons for you to show up:
The Mangia Pizza on Guadalupe is a lot smaller than it used to be, so please show up before 7:00 so we can have reasonable seating. Suggestions on new meeting locations are still welcome.
I'm repeating this information in case some of you didn't notice.
Old CACTUS passwords don't work on Bubba.cactus.org anymore. A new single-sign-on system is being installed and new temporary passwords had to be generated. It turns out that Apache and the Linux PAM modules support different password hash schemes, and re-using the old crypt password hashes was not an option.
So if you want to know what your new password is, please send me your contact information and I'll get that information to you as quickly as possible.
Over the past few years I've participated in several conversations where sys-admins express, sometimes quite colorfully, that their employer doesn't seem interested in providing resources for additional training. I've even done some of that expressing myself. It's clear that we have to take responsibility for our own professional development.
I've also participated in several discussions about declining CACTUS membership and the lack of compelling reasons for people to get in their cars and drive across town for a meeting. This is a common complaint for computer groups all over the country, that most everything you want to know about <your-OS-of-choice> is one or two Google searches away and that people are less willing to feed their inner geek on other people's schedules. Online forums are just too convenient and local computer clubs are dying like flies (or is that bees?).
I think I found an answer: The CACTUS HA-Cluster Project
Last year I started building a small high-availability cluster based on CentOS-5. I can't think of a better way to learn how clusters can fail other than building one, making it fail, and figuring out how to build it better.
Given the current cost of hardware, I can build a complete cluster for about $2,500. That would pay for one server for cluster management/storage (with 1-TB of disk) and three diskless machines that would be the cluster nodes (a minimal diskless node is around $500 with a 1U chassis). Compared to the cost of the RHCE Rapid Track Course and Exams ($3,600) this is a real bargain.
What I'd like to do, if there's sufficient interest, is to share the hardware I've already bought with CACTUS. If CACTUS is willing to help cover the costs, I'd be willing to relocate my hardware (and any hardware purchased by CACTUS) to one of our sponsors' co-location facilities and set it up so that CACTUS members can reserve time on it and use it for their own professional development. The only thing I'd ask is that members blog about what they're doing so that we can ALL learn from each other (and any mistakes that we might make).
Personally, I think something like this is totally consistent with CACTUS' history. CACTUS used to provide dialup Internet access and shell-accounts for its members when it was difficult to do that by yourself. Nobody needs that anymore. What IS difficult now, is assembling enough hardware to test really high-end enterprise configurations, like clusters.
Is anyone out there interested in doing hands-on work with clustering? Is this compelling enough for you to participate? To renew your membership, or join for the first time?
Comments and/or flames can be sent to my email address, which you will
find at the end of the newsletter.
p.s. The motherboard I'm using is the Supermicro X8SIL-F motherboard. It's a server-grade Micro-ATX board that, among other things, has an IPMI interface that allows for complete remote hardware management. This is an absolute necessity for running a Red Hat cluster. When a cluster node starts mis-behaving, the other nodes have to "remove" it before it causes damage. One way the cluster can do that is to contact the offending nodes' IPMI interface and issue a remote reset. This is known as STONITH: Shoot The Offending Node In The Head.
p.p.s. I plan on configuring the diskless nodes to boot to an iSCSI disk for the OS. I want to avoid the usual root-filesystem-over-NFS problems that one usually has with diskless configurations.
p.p.p.s. I do plan on experimenting with GFS and GFS2 on a SAN. Apparently, you don't have to have fibre-channel disks to do this, some people have implemented simple SAN fabrics using Co-Raid's ATA-over-Ethernet (AoE) drivers and AoE target software (see Wikipedia).
To renew your membership, please send check or money order payable to CACTUS ($30/yr for regular membership and $100/yr for corporate sponsorship):
PO BOX 9786
Austin, TX 78766-9786
You can also pay in person at the general meetings. Please direct any inquiries or address changes to membership [at] cactus <dot> org.
CACTUS meets on the third Thursday of each month. This month, the CACTUS meeting will be held at the angia Pizza on Guadalupe The address is 3016 Guadalupe Suite 100 (see http://tinyurl.com/ycsey5y). There is parking in the garage on the back.