Forum OpenACS Q&A: Which OS for new OpenACS.org

Collapse
Posted by Joel Aufrecht on
Mike Sisk reports that the original openacs.org server, which was down with a hardware problem, is repaired and ready for an OS.
"The next step is installing an OS; which flavor is up to the OCT.

For production machines we run Red Hat Enterprise 3 which costs $400 a
year. We could also go with Fedora; either Fedora Core 2 now, or wait a
week or two and see if Fedora Core 3 is released from beta.

Or we could go with something else like Debian, Suse, or FreeBSD; it's
up to you folks. I recommend a Red Hat distro only because it's well
supported and tested on Dell server hardware.

Here's a summary of the machine and its current configuration:

Dell PowerEdge 2550
1 1-GHz PIII CPU
1.5-GB RAM
4 36-GB SCSI drives in hardware RAID 5"
The OpenACS.org hosts in the past have run Red Hat.
Collapse
Posted by Malte Sussdorff on
I personally favour ubuntu (debian flavour) nowadays as it has been the most hassle free linux flavour I had the pleasure of installing so for.

I'm definitely against the RHLE3 solution due to the cost associated with it.

Kudos to Mike for all his work!

Collapse
Posted by Michael Hinds on
I like White Box. It's the free version of RHEL. I haven't used it in production yet, but it's behaving itself on my laptop.
Collapse
Posted by Talli Somekh on
Considering that RH is waiting for us to form some kind of organization so that they can donate the OACS copyright to us, I imagine we might be able to get them to donate a RHEL license.

That being said, if we're going all out free software, Debian or one of it's children would be neato.

(and might i point out that Ubuntu is maintained with GNU arch ;))

talli

Collapse
Posted by Roberto Mello on
I don't think it really matters these days if distro X or Z says it's more tested with Dells. They use standard hardware, and the kernel being used should be what metters.

Having said that, the distribution should be picked by who is going to be managing the box, whatever they feel is best. So I'd say the choice should be up to Mike.

And thanks again for the great work, Mike.

-Roberto

Collapse
Posted by Jade Rubick on
I'd vote for Debian stable with backports for newer packages. But I agree with Roberto that Mike should choose.
Collapse
Posted by Lamar Owen on
WhiteBox Enterprise Linux or any of the other RHEL 'shadows' would work fine and be well-supported by the Dell.  I am myself at PARI using WhiteBox in production, and it works well.  Yum updates, DAG's repository, etc, work together for me.

Having said that, I'm not running AOLserver on that box, so I don't know that portion of the equation.

What's this about RH donating copyrights?

Collapse
Posted by Andrew Piskorski on
Yet Another Distribution alternative is CentOS, which like White Box Linux, is a rebuild of the Red Hat Enterprise source rpms, with all non-open-source Red Hat content removed, etc. Why you'd pick one over the other, I've no idea.

The same folks also do cAos, which is a different rpm based distribution, not based on Red Hat. Personally, I may experiment with cAos at some point because the Warewulf cluster folks recommend it, but every time I look again at all these random different rpm-based distributions, I become annoyed and am again inclined to stick exclusively with Debian.

One little anecdote I say recently: Someone mention that their group had installed Debian on their Beowulf cluster - more than 7 years ago. Despite two major hardware upgrades, no re-installs at all, just lots of "apt-get upgrade".

I don't know of any other distribution that has actually maintained a solid upgrade path for 7+ years. Red Hat never even came close - even back in the Red Hat 6.x days c. 2000 or 2001, long before they orphaned all their non-Enterprise users, upgrading from one point release to another often didn't work anyway.

It is nice though to see many of the rpm-based distributions finally adopting apt-get or yum, and attempting to give Debian some competition.

Collapse
Posted by Matthew Geddert on
I would go with one of the "big ones" i.e. RedHat, FreeBSD, Suse, Debian and not one of the smaller distros though (even if they are based on the larger ones). As far as long term maintenance goes there would be less documenation online for nuanced differences between the niche's and the "big ones", and its easier to find volunteers that are really familiar with the bigger distros. As far as drivers and ease of install goes, as long as the hardware raid controller and Nic are supported by the distro you are set (we aren't installing a desktop here). What's the hardware raid controller for the server?

I would vote for Debian Sarge since it is very stable now, easy to maintain, and I simply like the community nature of the distro. Sarge can take the 2.6 kernel without difficulty (the upgrade for me from 2.4 to 2.6 was simply: apt get kernel-image...), although Debain Stable with backports would make sense too. 2.6 has been noticeably faster on my servers than 2.4 (although they do things other than just serve openacs sites, so i don't know if the differences are found in the other tasks).

Collapse
Posted by Andrew Piskorski on
Sarge is the next upcoming Debian 3.1 release. It hasn't been released to Stable yet, but it should be very close. Debian 3.0 (Woody) is so old now, and 3.1 is so close to being released, that I wouldn't even consider installing 3.0 on a server, I'd just install Unstable or Testing right now (not sure which), then switch over to Stable as soon as Sarge is released. I haven't actually tried that myself, but if Mike S. does decide to go with Debian, that's what I'd recommend.
Collapse
Posted by Jade Rubick on
The biggest problem with Sarge is that security updates are not supported for it.
Collapse
Posted by Joel Aufrecht on
See OCT OCT decision for Talli to ask RedHat to contribute RHEL license.
Collapse
Posted by russ m on
The biggest problem with Sarge is that security updates are not supported for it.
I seem to recall seeing recently (on debian-devel perhaps? I can't find a reference) that they're already putting together the infrastructure for security to support sarge before it gets shifted from testing to stable...
Collapse
Posted by russ m on
ah... here it is...
The bad news is that we still do not have an ETA for the testing-security autobuilders to be functional. This continues to be the major blocker for proceeding with the freeze; we would /like/ to have security support in place for sarge before encouraging widespread upgrade woody->sarge upgrade testing, but we /need/ to have it in place before releasing, so it would be unwise to try to freeze the rest of the archive without any confirmed schedule for the last stages of the release.
hmmm... and that message is over a month old... perhaps not so recent after all...