Greetings,
I *do* still have an FC2 box.
Would anyone second this procedure: http://www.centos.org/modules/newbb/viewtopic.php?topic_id=14052&forum=3...
Thanks.
Max Pyziur pyz@brama.com
On Fri, May 25, 2012 at 5:03 PM, Max Pyziur pyz@brama.com wrote:
I *do* still have an FC2 box.
Would anyone second this procedure: http://www.centos.org/modules/newbb/viewtopic.php?topic_id=14052&forum=3...
It might possibly work, but I can't quite imagine why anyone would want to do it at this point. Why not back up anything you might want to keep, install a nice clean Centos 6.x and put back the files you wanted?
On Fri, 25 May 2012, Les Mikesell wrote:
On Fri, May 25, 2012 at 5:03 PM, Max Pyziur pyz@brama.com wrote:
I *do* still have an FC2 box.
Would anyone second this procedure: http://www.centos.org/modules/newbb/viewtopic.php?topic_id=14052&forum=3...
It might possibly work, but I can't quite imagine why anyone would want to do it at this point. Why not back up anything you might want to keep, install a nice clean Centos 6.x and put back the files you wanted?
It's a test machine that replicates a production server. The production machine was setup in May 2011 when CentOS was in 5.8 and no 6.x had shown up.
So, I need a text 5.x box.
So do you (or anyone) second this or am I going to have to find out on my own and report back to you.
Max Pyziur pyz@brama.com
On Fri, May 25, 2012 at 7:42 PM, Max Pyziur pyz@brama.com wrote:
I *do* still have an FC2 box.
Would anyone second this procedure: http://www.centos.org/modules/newbb/viewtopic.php?topic_id=14052&forum=3...
It might possibly work, but I can't quite imagine why anyone would want to do it at this point. Why not back up anything you might want to keep, install a nice clean Centos 6.x and put back the files you wanted?
It's a test machine that replicates a production server. The production machine was setup in May 2011 when CentOS was in 5.8 and no 6.x had shown up.
So, I need a text 5.x box.
Even so, what's the point of an in-place upgrade compared to a fresh 5.x install? Even if it works, there will be old cruft left around that you don't need and that may cause surprises later.
On 05/25/2012 07:52 PM, Les Mikesell wrote:
On Fri, May 25, 2012 at 7:42 PM, Max Pyziur pyz@brama.com wrote:
I *do* still have an FC2 box.
Would anyone second this procedure: http://www.centos.org/modules/newbb/viewtopic.php?topic_id=14052&forum=3...
It might possibly work, but I can't quite imagine why anyone would want to do it at this point. Why not back up anything you might want to keep, install a nice clean Centos 6.x and put back the files you wanted?
It's a test machine that replicates a production server. The production machine was setup in May 2011 when CentOS was in 5.8 and no 6.x had shown up.
So, I need a text 5.x box.
Even so, what's the point of an in-place upgrade compared to a fresh 5.x install? Even if it works, there will be old cruft left around that you don't need and that may cause surprises later.
What Les said ...
If the production box is already CentOS 5.x ... it would seem to me that you already know what needs to be done to make your items run on CentOS-5.8.
If you upgrade a Fedora box to CentOS, while it can be done, it will contain many packages that are not part of CentOS. It will not be stable and it will not be a duplicate of your production box.
Backup the old info and wipe the machine, put 5.x on it, bring in the items you need from the backup (most of which you should know how to do, since you are already using it on 5.8 in production).
It is not worth the hassle of trying to remove all the Fedora Core items later on and doing an in-place upgrade ... at least not in my opinion.
On 05/25/2012 07:52 PM, Les Mikesell wrote:
On Fri, May 25, 2012 at 7:42 PM, Max Pyziur pyz@brama.com wrote:
I *do* still have an FC2 box.
Would anyone second this procedure: http://www.centos.org/modules/newbb/viewtopic.php?topic_id=14052&forum=3...
It might possibly work, but I can't quite imagine why anyone would want to do it at this point. Why not back up anything you might want to keep, install a nice clean Centos 6.x and put back the files you wanted?
It's a test machine that replicates a production server. The production machine was setup in May 2011 when CentOS was in 5.8 and no 6.x had shown up.
So, I need a text 5.x box.
Even so, what's the point of an in-place upgrade compared to a fresh 5.x install? Even if it works, there will be old cruft left around that you don't need and that may cause surprises later.
What Les said ...
If the production box is already CentOS 5.x ... it would seem to me that you already know what needs to be done to make your items run on CentOS-5.8.
If you upgrade a Fedora box to CentOS, while it can be done, it will contain many packages that are not part of CentOS. It will not be stable and it will not be a duplicate of your production box.
The point is to leave configurations, partitions, and other components as close as possible to being intact. Since this is a server environment, there are about 700-800 packages, not the 3000 that sit on desktop machine. Make lists of rpms on the FC2 install, and then sdiff'ing with the list of rpms installed from the CentOS upgrade should be one way of identifying non-CentOS packages and/or duplications.
Last, CentOS is built from Fedora Core 6. Usually, it makes sense to proceed sequentially. But how much difference is there from FC2 to FC6/CentOS 5.*?
MP pyz@brama.com
Backup the old info and wipe the machine, put 5.x on it, bring in the items you need from the backup (most of which you should know how to do, since you are already using it on 5.8 in production).
It is not worth the hassle of trying to remove all the Fedora Core items later on and doing an in-place upgrade ... at least not in my opinion.
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Wed, May 30, 2012 at 10:04 AM, Max Pyziur pyz@brama.com wrote:
The point is to leave configurations, partitions, and other components as close as possible to being intact.
Why isn't the point to match the existing CentOS box in production closely instead?
Since this is a server environment, there are about 700-800 packages, not the 3000 that sit on desktop machine.
If it is a server environment, you should be paying attention to the supported life of the distribution. FC2 is long, long past its 'use by' date.
Make lists of rpms on the FC2 install, and then sdiff'ing with the list of rpms installed from the CentOS upgrade should be one way of identifying non-CentOS packages and/or duplications.
Just get the package list from the working C5 box and feed it to kickstart or to yum after a minimal install.
Last, CentOS is built from Fedora Core 6. Usually, it makes sense to proceed sequentially.
No, it makes sense to upgrade things that were designed and tested as upgrades, and to re-install things that weren't. You might, with a lot of work and care, make the upgrade operational, but the result will be a one-of-a-kind beast that doesn't belong in a production environment.
But how much difference is there from FC2 to FC6/CentOS 5.*?
The point is that nobody knows, and there's no reason for anyone to know. You weren't supposed to run things that long on Fedora. But if you are going to let things go that long again with no maintenance, I'd recommend jumping all the way to C6 even if it is more work now, so 'yum update' will take care of it for years.
Les Mikesell wrote:
On Wed, May 30, 2012 at 10:04 AM, Max Pyziur pyz@brama.com wrote:
<snip>
Since this is a server environment, there are about 700-800 packages,
not the 3000
that sit on desktop machine.
If it is a server environment, you should be paying attention to the supported life of the distribution. FC2 is long, long past its 'use by' date.
Very much so. Almost anywhere I've ever worked, no management would *allow* a production server that was this far out of date.
Further, if it were up to me, there's *no* way I'd allow fedora in a production environment. It's a development line; I'd expect management to demand either RHEL or CentOS, which are stable production-quality lines. They don't have the latestgreatestmostwonderfulness... but when that moves into these distros, they're not going to break when you look at them wrong.
Just get the package list from the working C5 box and feed it to kickstart or to yum after a minimal install.
Last, CentOS is built from Fedora Core 6. Usually, it makes sense to proceed sequentially.
So you're going to upgrate to FC3, 4 and 5 before going to CentOS?
No, it makes sense to upgrade things that were designed and tested as upgrades, and to re-install things that weren't. You might, with a lot of work and care, make the upgrade operational, but the result will be a one-of-a-kind beast that doesn't belong in a production environment.
I agree. If someone handed me a mess like that, I'd be building a new production server, test it, and get that out of production as fast as I possibly could. If you, or whoever, got another job, or were hit by a car tomorrow, whoever had to pick it up would be SOL, and it'd probably crash before they figured out what had been done. It would take you as much time to document as to a) build a new, stable CentOS 5 or 6 box b) install everything required on it c) recompile anything in-house that needed to be rebuilt d) test it all, and put it into production,
and I guarantee that you'd miss documenting something vital.
But how much difference is there from FC2 to FC6/CentOS 5.*?
A *lot*. <snip> mark
Les Mikesell wrote:
On Wed, May 30, 2012 at 10:04 AM, Max Pyziur pyz@brama.com wrote:
<snip> >> Since this is a server environment, there are about 700-800 packages, not the 3000 >> that sit on desktop machine. > > If it is a server environment, you should be paying attention to the > supported life of the distribution. FC2 is long, long past its 'use > by' date.
Very much so. Almost anywhere I've ever worked, no management would *allow* a production server that was this far out of date.
Further, if it were up to me, there's *no* way I'd allow fedora in a production environment. It's a development line; I'd expect management to demand either RHEL or CentOS, which are stable production-quality lines. They don't have the latestgreatestmostwonderfulness... but when that moves into these distros, they're not going to break when you look at them wrong.
To clarify, the machine is a test/development box that also acts as a router to a DSL connection that (for the most part) replicates a co-located production machine that is currently running CentOS 5.8.
Until recently, energies have been dedicated to other endeavors. Currently, efforts are being made to upgrade all relevant components to appropriate recent stable releases of OS's. In no way was an FC2 machine used in a production environment, and no effort was made to create that impression.
Just get the package list from the working C5 box and feed it to kickstart or to yum after a minimal install.
Last, CentOS is built from Fedora Core 6. Usually, it makes sense to proceed sequentially.
So you're going to upgrate to FC3, 4 and 5 before going to CentOS?
Possibly. Unless someone else can attest to their own experience and knowledge that it's generally ok to move from FC2 to CentOS 5.*. That was my point in starting this thread.
MP pyz@brama.com
No, it makes sense to upgrade things that were designed and tested as upgrades, and to re-install things that weren't. You might, with a lot of work and care, make the upgrade operational, but the result will be a one-of-a-kind beast that doesn't belong in a production environment.
I agree. If someone handed me a mess like that, I'd be building a new production server, test it, and get that out of production as fast as I possibly could. If you, or whoever, got another job, or were hit by a car tomorrow, whoever had to pick it up would be SOL, and it'd probably crash before they figured out what had been done. It would take you as much time to document as to a) build a new, stable CentOS 5 or 6 box b) install everything required on it c) recompile anything in-house that needed to be rebuilt d) test it all, and put it into production,
and I guarantee that you'd miss documenting something vital.
But how much difference is there from FC2 to FC6/CentOS 5.*?
A *lot*.
<snip> mark
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Wed, May 30, 2012 at 11:07 AM, Max Pyziur pyz@brama.com wrote:
Last, CentOS is built from Fedora Core 6. Usually, it makes sense to proceed sequentially.
So you're going to upgrate to FC3, 4 and 5 before going to CentOS?
Possibly. Unless someone else can attest to their own experience and knowledge that it's generally ok to move from FC2 to CentOS 5.*. That was my point in starting this thread.
My experience with fedora was that a mid-rev update in FC5 included a kernel that would not run on the fairly mainstream IBM server where I was running it. So all bets are off...
Max Pyziur wrote:
Les Mikesell wrote:
On Wed, May 30, 2012 at 10:04 AM, Max Pyziur pyz@brama.com wrote:
<snip>
To clarify, the machine is a test/development box that also acts as a router to a DSL connection that (for the most part) replicates a co-located production machine that is currently running CentOS 5.8.
Until recently, energies have been dedicated to other endeavors. Currently, efforts are being made to upgrade all relevant components to appropriate recent stable releases of OS's. In no way was an FC2 machine used in a production environment, and no effort was made to create that impression.
Ok. That *was* the impression you gave. <snip>
Last, CentOS is built from Fedora Core 6. Usually, it makes sense to proceed sequentially.
So you're going to upgrate to FC3, 4 and 5 before going to CentOS?
Possibly. Unless someone else can attest to their own experience and knowledge that it's generally ok to move from FC2 to CentOS 5.*. That was my point in starting this thread.
*sigh* I was being sarcastic. Doing all that work would be silly, esp. with what would be needed to do so. Again, it would be *much* less work to build a good box of 5.8, or maybe 6.2, and load and configure that. <snip> mark
Max Pyziur wrote:
Les Mikesell wrote:
On Wed, May 30, 2012 at 10:04 AM, Max Pyziur pyz@brama.com wrote:
<snip> > To clarify, the machine is a test/development box that also acts as a > router to a DSL connection that (for the most part) replicates a > co-located production machine that is currently running CentOS 5.8. > > Until recently, energies have been dedicated to other endeavors. > Currently, efforts are being made to upgrade all relevant components to > appropriate recent stable releases of OS's. In no way was an FC2 machine > used in a production environment, and no effort was made to create that > impression.
Ok. That *was* the impression you gave.
No it wasn't. That was your mistaken interpretation.
<snip> >>>> Last, CentOS is built from Fedora Core 6. Usually, it makes sense to >>>> proceed sequentially. >> >> So you're going to upgrate to FC3, 4 and 5 before going to CentOS? > > Possibly. Unless someone else can attest to their own experience and > knowledge that it's generally ok to move from FC2 to CentOS 5.*. That > was > my point in starting this thread.
*sigh* I was being sarcastic. Doing all that work would be silly, esp.
You should do a better job of signalling your sarcasm.
with what would be needed to do so. Again, it would be *much* less work to build a good box of 5.8, or maybe 6.2, and load and configure that.
I'm not interested in acquiring more hardware but rather hope to use what I have. It works satisfactorily in its current configuration; my interest is in aligning the OS of the test/backup unit with that of the production machine.
<snip> mark
Max Pyziur pyz@brama.com
On 05/30/12 10:26 AM, Max Pyziur wrote:
I'm not interested in acquiring more hardware but rather hope to use what I have. It works satisfactorily in its current configuration; my interest is in aligning the OS of the test/backup unit with that of the production machine.
then back it up, wipe it and deploy 5.latest on the old hardware, reconfigure all your required services.
On 05/30/2012 12:37 PM, John R Pierce wrote:
On 05/30/12 10:26 AM, Max Pyziur wrote:
I'm not interested in acquiring more hardware but rather hope to use what I have. It works satisfactorily in its current configuration; my interest is in aligning the OS of the test/backup unit with that of the production machine.
then back it up, wipe it and deploy 5.latest on the old hardware, reconfigure all your required services.
exactly!
Max Pyziur wrote:
Max Pyziur wrote:
Les Mikesell wrote:
On Wed, May 30, 2012 at 10:04 AM, Max Pyziur pyz@brama.com wrote:
<snip> > To clarify, the machine is a test/development box that also acts as a > router to a DSL connection that (for the most part) replicates a > co-located production machine that is currently running CentOS 5.8. > > Until recently, energies have been dedicated to other endeavors. > Currently, efforts are being made to upgrade all relevant components to > appropriate recent stable releases of OS's. In no way was an FC2 > machine used in a production environment, and no effort was made to
create that
impression.
Ok. That *was* the impression you gave.
No it wasn't. That was your mistaken interpretation.
I accept that wasn't what you *intended*. However, what you *wrote* left that as a reasonable interpretation.
http://24.5-cent.us/egoless_documentation.doc
<snip>
Last, CentOS is built from Fedora Core 6. Usually, it makes sense to proceed sequentially.
So you're going to upgrate to FC3, 4 and 5 before going to CentOS?
Possibly. Unless someone else can attest to their own experience and knowledge that it's generally ok to move from FC2 to CentOS 5.*. That was my point in starting this thread.
*sigh* I was being sarcastic. Doing all that work would be silly, esp.
You should do a better job of signalling your sarcasm.
I did not expect you to actually consider that as within reason.
with what would be needed to do so. Again, it would be *much* less work to build a good box of 5.8, or maybe 6.2, and load and configure that.
I'm not interested in acquiring more hardware but rather hope to use what I have. It works satisfactorily in its current configuration; my interest is in aligning the OS of the test/backup unit with that of the production machine.
<snip>
Fine. Another answer would be to add more disk, if necessary, and build 5.8 on the machine, in such a manner as to allow you to reboot into either the current or the new version. For further clarification as to what I'm suggesting, try reading my other published article: http://24.5-cent.us/upgrading_linux.doc
mark
Max Pyziur wrote:
Max Pyziur wrote:
Les Mikesell wrote:
On Wed, May 30, 2012 at 10:04 AM, Max Pyziur pyz@brama.com wrote:
<snip> > To clarify, the machine is a test/development box that also acts as a > router to a DSL connection that (for the most part) replicates a > co-located production machine that is currently running CentOS 5.8. > > Until recently, energies have been dedicated to other endeavors. > Currently, efforts are being made to upgrade all relevant components > to > appropriate recent stable releases of OS's. In no way was an FC2 > machine used in a production environment, and no effort was made to
create that
impression.
Ok. That *was* the impression you gave.
No it wasn't. That was your mistaken interpretation.
I accept that wasn't what you *intended*. However, what you *wrote* left that as a reasonable interpretation.
Here is what I wrote: http://lists.centos.org/pipermail/centos/2012-May/126307.html
"... It's a test machine that replicates a production server. ..."
How would you improve it in order to remedy the apparent confusion?
http://24.5-cent.us/egoless_documentation.doc
<snip> >>>>>> Last, CentOS is built from Fedora Core 6. Usually, it makes sense >>>>>> to >>>>>> proceed sequentially. >>>> >>>> So you're going to upgrate to FC3, 4 and 5 before going to CentOS? >>> >>> Possibly. Unless someone else can attest to their own experience and >>> knowledge that it's generally ok to move from FC2 to CentOS 5.*. That >>> was my point in starting this thread. >> >> *sigh* I was being sarcastic. Doing all that work would be silly, esp. > > You should do a better job of signalling your sarcasm.
I did not expect you to actually consider that as within reason.
with what would be needed to do so. Again, it would be *much* less work to build a good box of 5.8, or maybe 6.2, and load and configure that.
I'm not interested in acquiring more hardware but rather hope to use what I have. It works satisfactorily in its current configuration; my interest is in aligning the OS of the test/backup unit with that of the production machine.
<snip>
Fine. Another answer would be to add more disk, if necessary, and build 5.8 on the machine, in such a manner as to allow you to reboot into either the current or the new version. For further clarification as to what I'm suggesting, try reading my other published article: http://24.5-cent.us/upgrading_linux.doc
Thanks. I've already looked at it.
I appreciate your and others' efforts at advice. I'm simply trying to use existing hardware (that's the eco-friendly approach), and trying to build my understanding of the Fedora/CentOS operational relationships. Given that it has been stated that CentOS 5.x was built from FC6, and that someone had already offered general guidance on the upgrade procedure (I shared the link in my initial request), I thought that it would be worthwhile asking the CentOS-users list to see if someone from this community had any direct experience with the upgrade. I'm not yet looking for a recommendation for a clean install.
Thanks.
Max Pyziur pyz@brama.com
mark
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Wed, May 30, 2012 at 1:21 PM, Max Pyziur pyz@brama.com wrote:
Here is what I wrote: http://lists.centos.org/pipermail/centos/2012-May/126307.html
"... It's a test machine that replicates a production server. ..."
How would you improve it in order to remedy the apparent confusion?
But in an earlier post you said it was a 'server environment' which at least sort-of implies that it is serving something.
I appreciate your and others' efforts at advice. I'm simply trying to use existing hardware (that's the eco-friendly approach), and trying to build my understanding of the Fedora/CentOS operational relationships.
Fedora doesn't support/recommend in-place upgrades across major versions or at least didn't for those versions. My experience was that even within a major rev. an update could kill your system. CentOS doesn't support/recommend in-place upgrades across major versions.
Given that it has been stated that CentOS 5.x was built from FC6, and that someone had already offered general guidance on the upgrade procedure (I shared the link in my initial request), I thought that it would be worthwhile asking the CentOS-users list to see if someone from this community had any direct experience with the upgrade. I'm not yet looking for a recommendation for a clean install.
I have seen success stories for FC6->CentOS conversions, along with some quirky stuff you have to to to fix it up. If you google enough you might be able to do that. However, FC2 was not at all like FC6 and I doubt if you'll find anyone who has made that or even a part of the FC2->FC6 path work. It would be crazy to try that without good backups. But if you have a place for the backups, you could use it instead to install and test a system that will work.
On Wed, May 30, 2012 at 1:21 PM, Max Pyziur pyz@brama.com wrote:
Here is what I wrote: http://lists.centos.org/pipermail/centos/2012-May/126307.html
"... It's a test machine that replicates a production server. ..."
How would you improve it in order to remedy the apparent confusion?
But in an earlier post you said it was a 'server environment' which at least sort-of implies that it is serving something.
The third post in the thread is the link that I cited above.
The first post in the thread (mine) - http://lists.centos.org/pipermail/centos/2012-May/126303.html - reads as follows
" ...Greetings,
I *do* still have an FC2 box.
Would anyone second this procedure: http://www.centos.org/modules/newbb/viewtopic.php?topic_id=14052&forum=3...
Thanks. ..."
In the course of the discussion, I did reference that it was a backup/test machine to a co-located CentOS box that is a production server.
But I clarified early in the thread that it was not production.
So how would you clarify the sentence of my second (and the third) posting of the thread so that it is unequivocally clear that it is not a production machine?
I appreciate your and others' efforts at advice. I'm simply trying to use existing hardware (that's the eco-friendly approach), and trying to build my understanding of the Fedora/CentOS operational relationships.
Fedora doesn't support/recommend in-place upgrades across major versions or at least didn't for those versions. My experience was that even within a major rev. an update could kill your system. CentOS doesn't support/recommend in-place upgrades across major versions.
That was made very clear in discussions following the introduction of CentOS 6.x.
Given that it has been stated that CentOS 5.x was built from FC6, and that someone had already offered general guidance on the upgrade procedure (I shared the link in my initial request), I thought that it would be worthwhile asking the CentOS-users list to see if someone from this community had any direct experience with the upgrade. I'm not yet looking for a recommendation for a clean install.
I have seen success stories for FC6->CentOS conversions, along with some quirky stuff you have to to to fix it up. If you google enough you might be able to do that. However, FC2 was not at all like FC6 and I doubt if you'll find anyone who has made that or even a part of the FC2->FC6 path work. It would be crazy to try that without good backups. But if you have a place for the backups, you could use it instead to install and test a system that will work.
Thank you; this is very helpful.
My hope is to upgrade; that way I don't have to change/specify partition topology, and hopefully only minimally adjust the existing configurations.
I have enough experience with unraveling rpm package dependency/duplication issues, having gone through F14->F15 DVD upgrade that failed/froze (in the end I worked with the "rescue" portion of the DVD and unraveled duplicate/missing package issues using yum and rpm; you can find that thread on the Fedora Users list).
-- Les Mikesell lesmikesell@gmail.com
MP pyz@brama.com
On Wed, May 30, 2012 at 1:56 PM, Max Pyziur pyz@brama.com wrote:
Thank you; this is very helpful.
My hope is to upgrade; that way I don't have to change/specify partition topology,
But that takes a couple of minutes - or you could use the saved kickstart info if FC2 saved it back then. Or install on the existing partitions.
and hopefully only minimally adjust the existing configurations.
And the part that matters should be in a dozen or so files in /etc. Save copies, diff/merge anything you don't understand well enough to do from scratch. Another few minutes.
I have enough experience with unraveling rpm package dependency/duplication issues, having gone through F14->F15 DVD upgrade that failed/froze (in the end I worked with the "rescue" portion of the DVD and unraveled duplicate/missing package issues using yum and rpm; you can find that thread on the Fedora Users list).
So you know that can take a long time to get right... I don't see the point of even considering it.
On Wednesday, May 30, 2012 02:56:24 PM Max Pyziur wrote:
My hope is to upgrade; that way I don't have to change/specify partition topology, and hopefully only minimally adjust the existing configurations.
I have tried this type of upgrade before; I have not had it go well for the most part. The only way I'd try to do an FC2 to C5 upgrade is by incrementally upgrading up to FC4 or FC5 using install media, then boot the C5.8 install media with 'upgradeany'. It may break things very badly.
I have had to do this sort of upgrade on SPARC systems running Aurora SPARC Linux; did a yum-based upgrade up through a few revs, and it was a pain. I only did it because install media wasn't already available, and you had to go backrev to get booting media on my particular box (although the installed system worked fine once installed). It is really something I would rather not do without the preupgrade logic in place, primarily because of non-repo or third-party repo packages that may or may not be around any more on a newer repo; for that matter, the Fedora package set in the FC2 days is likely to be larger than the C5 package set unless you enable third party repos at install/upgrade time, and that isn't guaranteed to work.
This sort of discussion is in the archives several times, and I think I have put my particular recipe out there before. It is recommended by the upstream vendor, Red Hat, to not do any major version upgrades from one version of EL to another. EL4 was based from around FC3, and you are essentially talking about a direct upgrade from a pre-EL4 to EL5; these two are more different than you might think. (see https://en.wikipedia.org/wiki/Red_Hat_Enterprise_Linux#Relationship_to_free_... for info)
Beyond that, the upgradeany path is probably the least tested of all the anaconda install paths, and will likely traceback at the worst possible time. Upgrades aren't easy (even on Debian/Ubuntu where packages being upgraded can ask questions and do significant things, unlike in the RPM scriptlet case). Preupgrade has failed for me more than it has worked, going through several revs of Fedora.
Having said all of that, if you analyze your particular package set and you figure out that all of the packages have identical configs between FC2 (or EL4, for that matter) and EL5, and that you're not using a package that has had major changes and upgrades break data (like PostgreSQL; FC2 shipped a significantly older PostgreSQL than CentOS 5 does, and a major version upgrade on PostgreSQL requires some special handling), you might be able to get it to work.
But it will probably take more time to successfully upgrade than it will to do a fresh install with the same list of packages and a restore of compatible configurations onto that fresh install. But, it's your time to waste if you want to do so.
If you want to see this sort of thing on the MS OS, there is a YouTube video out there highlighting upgrading through all versions of Windows; the cruft leftover from Window 1.0, 2.0, and 3.x in a Windows 7 upgraded system is a thing to behold.
On Wednesday, May 30, 2012 02:56:24 PM Max Pyziur wrote:
My hope is to upgrade; that way I don't have to change/specify partition topology, and hopefully only minimally adjust the existing configurations.
I have tried this type of upgrade before; I have not had it go well for the most part. The only way I'd try to do an FC2 to C5 upgrade is by incrementally upgrading up to FC4 or FC5 using install media, then boot the C5.8 install media with 'upgradeany'. It may break things very badly.
Just to advise the general readership. I downloaded iso's for FC3, FC4, FC5 DVD install discs, and their accompanying rescue CDs.
The machine under consideration is old by contemporary standards (a PIII-1400 w/ 1.5GB RAM, and three discs, one 2TB in size generally used to store backups.
The FC2->FC3->FC4-FC5 upgrades were done in about three hours; the time was split between checking the integrity of the DVDs and CDs and the upgrade. Today, I did the FC5->CentOS5.8 upgrade.
In each phase, the machine booted and functioned.
I recognized the postgresql issue you mention further in your posting; I've been through something like that several times, so I know how to work through it.
All-in-all, this has been easy; nothing like the FC14-FC15 DVD upgrade on my desktop that froze that I did two weeks ago (there, I spent a very large amount of time unraveling dependency issues and package duplications). I hope to do other FC upgrades in the spirit of being current, but I anticipate that it won't be as easy as the FC2 -> CentOS5.8 has been so far.
I recognize that most of the comments were from sysadmins, more involved in managing server farms, and steeped in that knowledge/experience base.
Much thanks to thoughtful comments and cautions,
fyi,
Max Pyziur pyz@brama.com
I have had to do this sort of upgrade on SPARC systems running Aurora SPARC Linux; did a yum-based upgrade up through a few revs, and it was a pain. I only did it because install media wasn't already available, and you had to go backrev to get booting media on my particular box (although the installed system worked fine once installed). It is really something I would rather not do without the preupgrade logic in place, primarily because of non-repo or third-party repo packages that may or may not be around any more on a newer repo; for that matter, the Fedora package set in the FC2 days is likely to be larger than the C5 package set unless you enable third party repos at install/upgrade time, and that isn't guaranteed to work.
This sort of discussion is in the archives several times, and I think I have put my particular recipe out there before. It is recommended by the upstream vendor, Red Hat, to not do any major version upgrades from one version of EL to another. EL4 was based from around FC3, and you are essentially talking about a direct upgrade from a pre-EL4 to EL5; these two are more different than you might think. (see https://en.wikipedia.org/wiki/Red_Hat_Enterprise_Linux#Relationship_to_free_... for info)
Beyond that, the upgradeany path is probably the least tested of all the anaconda install paths, and will likely traceback at the worst possible time. Upgrades aren't easy (even on Debian/Ubuntu where packages being upgraded can ask questions and do significant things, unlike in the RPM scriptlet case). Preupgrade has failed for me more than it has worked, going through several revs of Fedora.
Having said all of that, if you analyze your particular package set and you figure out that all of the packages have identical configs between FC2 (or EL4, for that matter) and EL5, and that you're not using a package that has had major changes and upgrades break data (like PostgreSQL; FC2 shipped a significantly older PostgreSQL than CentOS 5 does, and a major version upgrade on PostgreSQL requires some special handling), you might be able to get it to work.
But it will probably take more time to successfully upgrade than it will to do a fresh install with the same list of packages and a restore of compatible configurations onto that fresh install. But, it's your time to waste if you want to do so.
If you want to see this sort of thing on the MS OS, there is a YouTube video out there highlighting upgrading through all versions of Windows; the cruft leftover from Window 1.0, 2.0, and 3.x in a Windows 7 upgraded system is a thing to behold.
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Jun 3, 2012, at 12:55 PM, "Max Pyziur" pyz@brama.com wrote:
On Wednesday, May 30, 2012 02:56:24 PM Max Pyziur wrote:
My hope is to upgrade; that way I don't have to change/specify partition topology, and hopefully only minimally adjust the existing configurations.
I have tried this type of upgrade before; I have not had it go well for the most part. The only way I'd try to do an FC2 to C5 upgrade is by incrementally upgrading up to FC4 or FC5 using install media, then boot the C5.8 install media with 'upgradeany'. It may break things very badly.
Just to advise the general readership. I downloaded iso's for FC3, FC4, FC5 DVD install discs, and their accompanying rescue CDs.
The machine under consideration is old by contemporary standards (a PIII-1400 w/ 1.5GB RAM, and three discs, one 2TB in size generally used to store backups.
The FC2->FC3->FC4-FC5 upgrades were done in about three hours; the time was split between checking the integrity of the DVDs and CDs and the upgrade. Today, I did the FC5->CentOS5.8 upgrade.
In each phase, the machine booted and functioned.
I recognized the postgresql issue you mention further in your posting; I've been through something like that several times, so I know how to work through it.
All-in-all, this has been easy; nothing like the FC14-FC15 DVD upgrade on my desktop that froze that I did two weeks ago (there, I spent a very large amount of time unraveling dependency issues and package duplications). I hope to do other FC upgrades in the spirit of being current, but I anticipate that it won't be as easy as the FC2 -> CentOS5.8 has been so far.
I recognize that most of the comments were from sysadmins, more involved in managing server farms, and steeped in that knowledge/experience base.
Much thanks to thoughtful comments and cautions,
You might want to crawl /etc, /bin, /sbin, /usr and /var for files not under management and see if you have anything left over.
I might use find here,
# find /etc -type f -exec rpm -qf {} ; -print
Of course the rpm command should be tweaked so it it just returns an error code if the file isn't in the database instead of any output and have find -print the path so you can redirect the output to a file.
Remember not all files not under management are orphan files, so you will need to use some knowledge to figure out which you can rm.
-Ross
On Sun, 3 Jun 2012, Ross Walker wrote:
On Jun 3, 2012, at 12:55 PM, "Max Pyziur" pyz@brama.com wrote:
[... deleted for the sake of brevity ...]
Much thanks to thoughtful comments and cautions,
You might want to crawl /etc, /bin, /sbin, /usr and /var for files not under management and see if you have anything left over.
I might use find here,
# find /etc -type f -exec rpm -qf {} ; -print
Of course the rpm command should be tweaked so it it just returns an error code if the file isn't in the database instead of any output and have find -print the path so you can redirect the output to a file.
Remember not all files not under management are orphan files, so you will need to use some knowledge to figure out which you can rm.
Great advice, and I'll take it up shortly.
Much thanks.
MP pyz@brama.com
-Ross
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Sun, Jun 3, 2012 at 11:55 AM, Max Pyziur pyz@brama.com wrote:
I recognize that most of the comments were from sysadmins, more involved in managing server farms, and steeped in that knowledge/experience base.
And in upgrades, and thus have experience with the difference in effort and results....
And you still have yet to explain what you think the result of the extra hours of downloading and work has gained compared to a clean install and copying your data back. As far as I can see, it is a bunch of orphaned files, a wildly fragmented disk layout, and probably a less efficient filesystem. This is especially true since you mentioned having several disks, one probably large enough to hold a complete backup of the old system disk so you could easily pick out what you want back in the new install.
On Sun, 3 Jun 2012, Les Mikesell wrote:
On Sun, Jun 3, 2012 at 11:55 AM, Max Pyziur pyz@brama.com wrote:
I recognize that most of the comments were from sysadmins, more involved in managing server farms, and steeped in that knowledge/experience base.
And in upgrades, and thus have experience with the difference in effort and results....
My point was that it is a different focus, knowledge, and experience base. If server farm system administration is what you know, then you will place all issues in that framework. If all you know how to use is a jackhammer, then you'll approach every problem in the same way.
My personal goal was to preserve the topology of the disk layout, as well as the configurations.
And you still have yet to explain what you think the result of the extra hours of downloading and work has gained compared to a clean
What extra hours? The downloading was done overnight, the diskburning was relatively quick.
install and copying your data back. As far as I can see, it is a bunch of orphaned files, a wildly fragmented disk layout, and probably
This is now an example of a casebook fallacy - a strawman argument. W/o investigating anything, you've projected a set of unsubstantiated qualifications on a situation and are now arguing against them.
a less efficient filesystem. This is especially true since you mentioned having several disks, one probably large enough to hold a complete backup of the old system disk so you could easily pick out what you want back in the new install.
The 2TB disk is where backups for several resident machines resides (notebooks, desktops); it's about 84% full. Admittedly, there's space for configurations, but that was not my first interest (I think that I've stated that at least twice).
MP pyz@brama.com
On Sun, Jun 3, 2012 at 1:53 PM, Max Pyziur pyz@brama.com wrote:
My personal goal was to preserve the topology of the disk layout, as well as the configurations.
Which are trivial to reproduce. And potentially improve in the process.
And you still have yet to explain what you think the result of the extra hours of downloading and work has gained compared to a clean
What extra hours? The downloading was done overnight, the diskburning was relatively quick.
You did mention several intermediate versions that would not have been needed. You can't possibly claim it did not take extra time.
install and copying your data back. As far as I can see, it is a bunch of orphaned files, a wildly fragmented disk layout, and probably
This is now an example of a casebook fallacy - a strawman argument. W/o investigating anything, you've projected a set of unsubstantiated qualifications on a situation and are now arguing against them.
No, I'm asking what you think you gained, and you still can't describe how your result is an improvement over a fresh install.
The 2TB disk is where backups for several resident machines resides (notebooks, desktops); it's about 84% full. Admittedly, there's space for configurations, but that was not my first interest (I think that I've stated that at least twice).
Yes, you did say that, but why? Did you just want to prove it can be done the hard way, or do you think your machine is somehow better now?
On Sun, 3 Jun 2012, Les Mikesell wrote:
On Sun, Jun 3, 2012 at 1:53 PM, Max Pyziur pyz@brama.com wrote:
My personal goal was to preserve the topology of the disk layout, as well as the configurations.
Which are trivial to reproduce. And potentially improve in the process.
It may be trivial for you, but not for me.
And you still have yet to explain what you think the result of the extra hours of downloading and work has gained compared to a clean
What extra hours? The downloading was done overnight, the diskburning was relatively quick.
You did mention several intermediate versions that would not have been needed. You can't possibly claim it did not take extra time.
I measured the amount of time that it took against a recent FC14->FC15 upgrade via DVD. That took the bulk of a weekend.
This series of sequential upgrades took a few hours.
install and copying your data back. As far as I can see, it is a bunch of orphaned files, a wildly fragmented disk layout, and probably
This is now an example of a casebook fallacy - a strawman argument. W/o investigating anything, you've projected a set of unsubstantiated qualifications on a situation and are now arguing against them.
No, I'm asking what you think you gained, and you still can't describe how your result is an improvement over a fresh install.
Ok, I sense that there is some sort of affront and that you need to defend yourself. I'm not challenging you, your experience, your capabilities, and your knowledge. They are commendable.
Nevertheless, what several individuals, including yourself, have presented is that a fresh install is the optimal solution; anything else has elevated risk. In my case, I've preserved my disk topology and my configurations, which was one of my priorities; I've expended a few hours, and I'm doing other things.
The 2TB disk is where backups for several resident machines resides (notebooks, desktops); it's about 84% full. Admittedly, there's space for configurations, but that was not my first interest (I think that I've stated that at least twice).
Yes, you did say that, but why? Did you just want to prove it can be done the hard way, or do you think your machine is somehow better now?
Never said this is the hard way; but it definitely is not that challenging, especially since I don't have the experience or knowledge of yeoman sysadmins.
And I'm not seeking to win a trophy for my machine; I'm seeking basic and simple continuity.
fyi,
MP pyz@brama.com
On 5/30/2012 2:21 PM, Max Pyziur wrote:
I appreciate your and others' efforts at advice. I'm simply trying to use existing hardware (that's the eco-friendly approach), and trying to build my understanding of the Fedora/CentOS operational relationships. Given that it has been stated that CentOS 5.x was built from FC6, and that someone had already offered general guidance on the upgrade procedure (I shared the link in my initial request), I thought that it would be worthwhile asking the CentOS-users list to see if someone from this community had any direct experience with the upgrade. I'm not yet looking for a recommendation for a clean install.
The issue here is that upgrading between major versions (CentOS 4.x to 5.x) is not supported or recommended. And if it is not recommended to upgrade from CentOS 4 to CentOS 5, it is even less recommended to try an upgrade from Fedora Core 2 to CentOS 5.
As has been mentioned before, the results of the upgrade will be a box that claims to be CentOS, but has lots of extraneous packages and files left behind from the previous Fedora install. This will result in a system that seems to run fine, but is likely to have strange problems from time to time when something tries to use one of these old files.
On 05/30/12 11:21 AM, Max Pyziur wrote:
I'm not yet looking for a recommendation for a clean install.
there is no other sane approach. you can upgrade the system a half dozen times through those intermediate versions, and sort out every issue that comes along, or you can build a clean new system (either on the same hardware or not) and do it all at once.
this is a test/dev box for an existing EL5 production system? make a backup of the production system and restore it on the dev box, reconfigure the network, yum install any additional development packages (C compilers, etc), done, go home and have a beer.
Max Pyziur wrote:
Max Pyziur wrote:
Max Pyziur wrote:
Les Mikesell wrote: > On Wed, May 30, 2012 at 10:04 AM, Max Pyziur pyz@brama.com wrote:
<snip>
I appreciate your and others' efforts at advice. I'm simply trying to use existing hardware (that's the eco-friendly approach), and trying to build my understanding of the Fedora/CentOS operational relationships. Given that it has been stated that CentOS 5.x was built from FC6, and that someone had already offered general guidance on the upgrade procedure (I shared the link in my initial request), I thought that it would be worthwhile asking the CentOS-users list to see if someone from this community had any direct experience with the upgrade. I'm not yet looking for a recommendation for a clean install.
And we *all* are saying that doing other than a clean install, or a parallel install, as I suggested in the article, is a bad idea.
mark
On 05/30/12 9:07 AM, Max Pyziur wrote:
Possibly. Unless someone else can attest to their own experience and knowledge that it's generally ok to move from FC2 to CentOS 5.*. That was my point in starting this thread.
sure. take new system, clean install 5.latest on it, configure your services. deploy, retire/recycle old box.
-
Bowie Bailey -
John R Pierce -
Johnny Hughes -
Lamar Owen -
Les Mikesell -
m.roth＠5-cent.us -
Max Pyziur -
Ross Walker