Hello all, After working with Linux on a number of levels for the last 5 years, I've decided to try FreeBSD. I just finished installing 4.9, and I have everything tweaked to fit my needs. I'm tracking RELENG_4_9, and so far so good with everything. Now on to my question, I just noticed that 5.3 will probably be coming out sometime in the near future, which makes me wonder if I've made the wrong decision in installing 4.9. Honestly, there aren't really any features in 5.3 that I think I'll really need, but I'm concerned about the lifespan of the 4.x branch. How long will it be until ports start failing due to using an older branch? How long can I feasibly run on the 4.x branch? I wouldn't be so concerned; however, once this server is in place, it will be running some very critical services and the thought of backing everything up and installing from scratch to migrate to 5.x just represents too much unnecessary downtime. It seems that there isn't any clean way to upgrade between major versions due to differing filesystems (UFS and UFS2) and leftover relics from previous releases causing potential problems. What do most people do for their longterm upgrade strategy with FreeBSD on production servers? Sorry for the long-winded question, but I want to get this right the first time, so I don't end up kicking myself down the road. I'd also prefer not to have to do a full reinstall once a year or whenever a new major release happens just to get support. One last thing, I recently patched a port, dbf2mysql, in order to get it to work properly with mysql323-server. What is the proper way to get this patch committed to the ports tree? It seems generally useful to anyone using the program, and I would rather not have to manually patch everytime I upgrade the port. Thanks, Travis Whitton
Travis Whitton wrote:> Hello all, > After working with Linux on a number of levels for the last 5 years, I've > decided to try FreeBSD.Woo hoo, welcome to the fun!> I just finished installing 4.9, and I have everything > tweaked to fit my needs. I'm tracking RELENG_4_9, and so far so good with > everything.Good news.> Now on to my question, I just noticed that 5.3 will probably be coming out > sometime in the near future, which makes me wonder if I've made the wrong > decision in installing 4.9. Honestly, there aren't really any features in 5.3 > that I think I'll really need, but I'm concerned about the lifespan of the 4.x > branch. How long will it be until ports start failing due to using an older > branch? How long can I feasibly run on the 4.x branch?Ports shouldn't fail at all for a year or more. You didn't mention what kind of hardware this is. If it's a single processor system, 4.9 is probably going to give you the best performance in that one year time frame. If it's an SMP system, in the next 6 months or so 5.x will be a significant benefit.> I wouldn't be so concerned; however, once this server is in place, it will be > running some very critical services and the thought of backing everything up > and installing from scratch to migrate to 5.x just represents too much > unnecessary downtime. It seems that there isn't any clean way to upgrade > between major versions due to differing filesystems (UFS and UFS2) and leftover > relics from previous releases causing potential problems.No, there isn't, and particularly with the issue of new file systems, there cannot be. Your best bet is to do the backup and restore, as you mentioned. That said, I'd like to suggest an alternative proposal. If this system is so crucial to your operation, it ought to have some redundancy, right? :) Why not do a head to head test with 4.9 and 5.x-current on the same hardware? That way you solve several problems at the same time. You'll be able to determine conclusively if 5.x works for you, you'll have a hot spare system ready to go in case of a hardware failure, and you won't have any downtime at all during upgrade cycles.> One last thing, I recently patched a port, dbf2mysql, in order to get it to > work properly with mysql323-server. What is the proper way to get this patch > committed to the ports tree? It seems generally useful to anyone using the > program, and I would rather not have to manually patch everytime I upgrade > the port.man send-pr Good luck, Doug -- This .signature sanitized for your protection
> Ports shouldn't fail at all for a year or more. You didn't mention what > kind of hardware this is. If it's a single processor system, 4.9 is > probably going to give you the best performance in that one year time > frame. If it's an SMP system, in the next 6 months or so 5.x will be a > significant benefit.It's a single proc system. A year seems long enough to merit sticking with 4.9. It seems that major releases can often stretch to over a year, so that should give me a while to stay on the 5.x tree once I decide it's safe to switch.> That said, I'd like to suggest an alternative proposal. If this system > is so crucial to your operation, it ought to have some redundancy, > right? :) Why not do a head to head test with 4.9 and 5.x-current on > the same hardware? That way you solve several problems at the same time. > You'll be able to determine conclusively if 5.x works for you, you'll > have a hot spare system ready to go in case of a hardware failure, and > you won't have any downtime at all during upgrade cycles.Not a bad idea at all. We do regular backups of course, but it's always nice to have a drop-in replacement. Thanks very much for your input. I look forward to being a member of the outstanding FreeBSD community. FYI, the server will be acting as a repository and access point to many years of historical(and current) environmental data, so you can add that to the list of tasks that FreeBSD performs. Thanks for your help, Travis Whitton