Fedora 17 Production Server

[koffiejunkie]

^^This - and I'm a Fedora adherent. Put CentOS on your production server and experiment on your dev systems rather.

This reminds me of one of my pet peeves. Developers using the latest Fedora/Ubuntu/Gentoo/Mint/Arch/whatever to develop and test their software, and then come crying to me when it doesn't run properly on their RHEL servers. Idiots.

 

[w1z4rd]

My choice:

CentOS for web hosting servers.
Debian for running specific applications
BSD (for ZFS) for backup servers

 

[koffiejunkie]

CentOS for web hosting servers.
Debian for running specific applications

Odd, I have them the other way around. Well, at work we use RHEL, so I have no choice, but I prefer Debian for small solutions for most applications. RHEL/CentOS still doesn't have a way to make Apache speak to php-fpm out of the box, and probably won't in RHEL7/CentOS7 either. Of course, at work I build the packages for this, but it means I have to maintain it, and I can't complain to our Red Hat account manager if I stumble upon bugs.

BSD (for ZFS) for backup servers

I'm liking that idea too. I've tried the two solutions for ZFS under linux (Fuse and zfsonlinux) - I don't recommend either for anything more than tinkering with it to get to know it. It's really flakey.

 

[debonair]

I am yet to try BSD... will download an ISO and tinker with it.. which flavor do you guys suggest, the ones with the most resources?

 

[koffiejunkie]

I would stick with FreeBSD if I was just getting into it. It has the biggest community, and excellent documentation:

http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/

 

[ponder]

Seeing you guys are talking about BSD & ZFS see,

http://bsdmag.org/magazine/1809-tuning-zfs-on-freebsd

Tuning ZFS on FreeBSD

ZFS is a modern 128-bit file system based on the copy-on-write model. It originates from the OpenSolaris project and has first appeared in FreeBSD in 2008. ZFS has many innovative features including an integrated volume manager with mirroring and RAID capabilities, data checksumming and compression, writable snapshots that can be transferred between systems and many more. In this article the author is going to discuss several tuning options including sysctl(2) knobs and give examples how can ZFS performance and efficiency can be measured and evaluated. This article is intended for FreeBSD users with ZFS version 28 available since 8.3-RELEASE and 9.0-RELEASE.

 

[volstruis]

CentOS is OK I have used it before on the amazon cloud. If I do experience show stopping issues I may consider it. The problem with most sysadmins is that you guys don't like to experiment. Distributions like Fedora can be very stable, they go through rigorous testing from people. If you take the time to learn whats going on you can configure a really solid environment and learn a thing or 2 in the process.

A distribution shows what its really made of when it is stress tested under the load of a production environment.

This shows you haven't had any real world experience with production environments.
There is penalties payable if you cause down time to the big boys systems. Think R50k per 10mins at the lighter end of the scale. This will get you fired I have seen it happen more than once. So I think its actually the other way round that you need to take time and learn what is actually going on in the field.

 

[koffiejunkie]

There is penalties payable if you cause down time to the big boys systems. Think R50k per 10mins at the lighter end of the scale.

While that is true in principle, it's also BS 99% of the time. I see this every day at work (I work for a hosting co that hosts these "big boys" but also the small boys and everything in-between): Client has downtime for whatever reason, and scream at us that they're losing tens of thousands of dollars per hour or millions a day. But they're hosting their entire business on a single server.

The companies who are serious about keeping their stuff running, choose the right hardware to facilitate this, and build their applications to cope with the loss of a node or multiple nodes, and some actively break stuff in their production environments - Netflix is famous for this.

 

[debonair]

This shows you haven't had any real world experience with production environments.
There is penalties payable if you cause down time to the big boys systems. Think R50k per 10mins at the lighter end of the scale. This will get you fired I have seen it happen more than once. So I think its actually the other way round that you need to take time and learn what is actually going on in the field.

lol I guess I dont