Jun Martin <[EMAIL PROTECTED]> writes: > Ambrosio Berdijo wrote: >> Lastly, some reasons why people still run old Linux/BSD/Unix systems. >> >> 1. Why change when everything is running well. Why risk upgrading to >> the latest/greatest kernel when firewall/router/dns/email system is >> already secured and running well. It may even be riskier migrating >> all the settings+data to a new system. > > You hit the nail on the head right there!
Thinking simply, this is a good assertion. However: - The OT was talking of building packages needing a new GLibc. That obviously means that there's an upgrade in consideration, so it's the OT's job to assess the risks (which he does by asking this list, which is a good thing to do :D) and decide whether to pursue upgrading the components or not. - The fact that GLibc is the component being considered for upgrading presents its own problems. Since many systems consider the C library to be the foremost component needed for userspace programs to work (not to mention some kernelspace utilities as well,) one has to really weigh in the pros and cons of upgrading such a sensitive component. For obvious reasons, one has to be really intimate with the internals of the system being considered to be able to fully assess the worth of an upgrade. Development servers keep upgrading to the latest and the greatest because the developers using it are keen on the new developments and are (usually) aware that changes will probably cause data loss, if not careful. Production servers, on the other hand, keep upgrading mainly to avail of security updates, and not because of new functionality. And for systems with active support contracts, upgrading is necessary because its what you paid support for. You could possibly nullify a support contract by not upgrading, or doing a partial upgrade. >> 2. You have a lot of data on your system. Ask yourself, do you really >> want to take the risk of losing all the StarGate/StarTrek episodes >> stored in your RAID5 ReiserFS disk system just because of a new >> RedHat release? > > Backup/Restore would sold this, but if it's a production > server.......... well, you know... Whether it's SG-1, or Atlantis, or git trees, or a database, you probably should be putting those in some archival storage dump in the first place. And no, RAID isn't a dumpfs... >> 3. You have some major software (e.g. Oracle database) that >> specifically works on that version of the distro. > > 4. Your hardware is certified to run only on that version of that > distro, and the manufacturer has included CD's for that distro version > only and they support only that distro version for that hardware... Worst cases like this can be mitigated, although with some effort, by running the old version of the system within the newer version. There are plenty ways (stupid chroot, qemu, kvm, xen, and lguest, to name a few) one can do about it. You obviously won't be thinking about calling support when doing this. OTOH, it shouldn't(?) be too hard to just call up support and ask if you can avail of a new version of that major software, and maybe a recertification or two on the hardware, plus a few more add-ons. Of course there's cost, but I'll assume you have covered that. Cheers, Zakame -- Zak B. Elep [EMAIL PROTECTED] _________________________________________________ Philippine Linux Users' Group (PLUG) Mailing List [email protected] (#PLUG @ irc.free.net.ph) Read the Guidelines: http://linux.org.ph/lists Searchable Archives: http://archives.free.net.ph
