Major Version Upgrade 4.11 to 5.x

Matthew Seaman m.seaman at infracaninophile.co.uk
Wed Dec 13 03:27:55 PST 2006


Lane wrote:
> On Monday 11 December 2006 01:18, Matthew Seaman wrote:
>> listvj wrote:
>>> I'm interested in upgrading from 4.11 to 5.x.  I currently track 4.x
>>> stable using cvsup, but I've never done a major version upgrade.
>>>
>>> First, should I bother?  My hardware has dual pentium 1.13 processors
>>> with 1G ram (I'm considering maxing it out at 4).  I host email and web
>>> sites for a few domains on this machine and I have four jails configured
>>> on it which will have to be upgraded too.  I have users counting
>>> particularly on mail service not being down for too long.
>>>
>>> Other than the obvious advice to start with a good backup, can anyone
>>> tell me:
>>>
>>> 1)  Will I gain a major benefit from upgrading
>>> 2)  Where should I look for instructions / advice on upgrading
>>> 3)  Also any general advice from personal experience.
>>> 4)  Just how risky is this?
>> Uh -- why upgrade to a branch (5.x) that has already had it's last
>> release and is worse performing than both 4.x and 6.x?  You should
>> really be looking at upgrading to 6.2-RELEASE just as soon as it
>> comes out (Real Soon Now).
>>
>> As for risk -- for various reasons you will be better off doing a
>> clean install of 6.x and rebuilding your server from the ground up.
>> It's no more risky than installing any other server -- unless you
>> have some legacy binary-only application that you absolutely have
>> to run, it is virtually certain to succeed.
>>
>> You biggest problem would seem to be the downtime required to do
>> the update -- if you can manage it, probably the least consumer
>> impact method is building the upgraded system on fresh disks on a
>> scratch box, and then finishing the upgrade by a disk-swap.  Which
>> also has the added benefit that you have a ready-made back out
>> path.
>>
>> 	Cheers,
>>
>> 	Matthew
> Matthew,
> 
> I agree with your advice to build the new server with a clean install, if only 
> to prevent any sendmail issues.
> 
> But I'm not so sure I understand your assessment that 5.x is "worse performing 
> than both 4.x and 6.x."  While I agree that 6.x is a great improvement in 
> functionality over 5.x, I was not aware of the poor performance record of 
> 5.x.  
> 
> Do you know of any links to benchmark tests, or other data, which would 
> provide some more background on this?
> 
> That kind of data would greatly influence my opinion in this discussion.  
> Without it I'd be pleased to recommend 5.X, regardless of it's pending "drop 
> dead" date, wrt support.  I certainly see no need to chain myself to any 
> software release cycle, nor, it seems, does the original poster.  I'm in awe 
> of his patience, and clearly he is satisfied with the product if he remains 
> on 4.11.
> 
> Thanks,
> 
> lane
> ~Still running 5.x

That's comment was based on my experience running a few hundred FreeBSD
servers of various models and OS versions.  I should qualify that by saying
that 4.x performance really shines when you're using single processor boxes
and not running heavily multithreaded applications.  On the other hand,
6.x does very well all round, especially with multithreaded applications
and multiple CPUs.  Of course, you also need 6.x for AMD64 support.

5.x wasn't in any sense bad, but the difference in performance between
5.x and 6.x is very obvious even without running exhaustive benchmarks.  
There's no good reason I know of to prefer 5.x to 6.x.

Remember too that the policy about when releases were created and how
they were numbered changed between 5.x and 6.x: previously a major version
number change was made when some target set of functionality was
implemented.  Now the major version number is bumped every 18 months (I
think -- something like that anyhow), using whatever new stuff has gone
into HEAD since the last major bump.  6.x is in many ways what the project
had intended 5.x to be, before becoming mired in the difficult transition
from 4.x to 5.x.  

	Cheers,

	Matthew

-- 
Dr Matthew J Seaman MA, D.Phil.                       Flat 3
                                                      7 Priory Courtyard
PGP: http://www.infracaninophile.co.uk/pgpkey         Ramsgate
                                                      Kent, CT11 9PW, UK

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 250 bytes
Desc: OpenPGP digital signature
Url : http://lists.freebsd.org/pipermail/freebsd-questions/attachments/20061213/73c43fda/signature-0001.pgp


More information about the freebsd-questions mailing list