Hi,
I've been contacted by a local training center specialized in Oracle databases, to train a group of four administrators to use Linux. They're supposed to use Oracle Linux (more exactly "OL5"), which I understand is some specialized version of Red Hat Enterprise Linux.
I'm reasonably proficient with CentOS. I've been using it exclusively on desktops and servers since 2007 (version 4.3 if I remember correctly), after a few years on Slackware and Debian. I've setup CentOS-based networks in small town halls, public libraries and schools, my dedicated webserver is running the latest CentOS, and I've published a book about basic Linux concepts entirely based on CentOS 5.3.
On the other hand, I've never got to work on the "real thing", so to speak. The only time I put my hands on a Red Hat Enterprise Linux server was to retrieve the root password for a distracted client (add init=/bin/bash to the kernel line in GRUB, mount -o remount,rw /, passwd :oD). And I never came near a machine running Oracle Linux.
What differences can I expect between a server running Oracle Linux and my average server running CentOS ? As far as I can guess, their "unbreakable" Linux kernel will still be some package called 'kernel' (with the according 'kernel-devel' and 'kernel-headers' packages). Probably Yum will be used as a dependency resolver (will it?), only with different repositories. And for the rest, I except it to work the same, in that I can still use chkconfig, ntsysv, rpm, etc. (what about system-config-securitylevel-tui or similar tools?)
In short, what differences/similarities can I expect? Is there some document stating these?
Cheers from the sunny South of France,
Niki Kovacs
Hi,
I've been contacted by a local training center specialized in Oracle databases, to train a group of four administrators to use Linux. They're supposed to use Oracle Linux (more exactly "OL5"), which I understand is some specialized version of Red Hat Enterprise Linux.
I think they offer you a choice of kernels. Their own kernel has performance enhancements.
It's also supposed to be cheaper in maintenance than RHEL. But AFAIK, it's a re-engineering of RHEL in the same way CentOS is. Just with commercial support (and geared towards running Oracle stuff).
I just looked and at the most basic level (patches only) it costs 94/year/system.
Does anybody know how fast they implement the fixes from upstream?
Rainer
On Wed, Nov 24, 2010 at 09:50:07AM +0100, Niki Kovacs wrote:
Hi,
What differences can I expect between a server running Oracle Linux and my average server running CentOS ? As far as I can guess, their "unbreakable" Linux kernel will still be some package called 'kernel' (with the according 'kernel-devel' and 'kernel-headers' packages). Probably Yum will be used as a dependency resolver (will it?), only with different repositories. And for the rest, I except it to work the same, in that I can still use chkconfig, ntsysv, rpm, etc. (what about system-config-securitylevel-tui or similar tools?)
Well, it's pretty similar. I'm not sure if I messed up the installation or not, but I had to later download reops--it has yum, chkconfig, service and so on.
I didn't find a samba package, but as this was for one oracle developer to transfer files, we just used winscp. (I also didn't look hard for a samba package, at present, this is just a test.)
Disk naming seems slightly different--we went with more standard partitions rather than LVM, and rather than /dev/sda and so on, it was /dev---arrgh, I'm not at work, and my mind just went blank--errm, xda and xdb? At any rate, it will be fairly obvious.
As mentioned, this is a test, and while I don't know all the circumstances, we're not entitled to support--if you do it officially, as opposed to what is probably a trial version, there is support, and I assume (but note, that's an assumption, not knowledge), one could call and ask things like, Where can I find samba?
(Samba client and samba common was available, but no smbd. Again, I did NO googling on this, just did yum provides */smbd, got no results, and then used WinSCP.)
There won't be any major shocks. On the other hand, in our case, there's nothing at all tricky for my part of it--I just had to get a basic server running for the oracle developers to use.
On 24 November 2010 11:57, Scott Robbins scottro@nyc.rr.com wrote:
Well, it's pretty similar. I'm not sure if I messed up the installation
It's almost identical. Oracle kernel parameters & dependencies can be automatically installed during installation albeit later on Oracle RDBMS installation gives you warnings since it doesn't like the default values in /etc/sysctl.conf as populated by OEL.
The rest? Almost completely identical to upstream.
I didn't find a samba package, but as this was for one oracle developer to transfer files, we just used winscp. (I also didn't look hard for a samba package, at present, this is just a test.)
Are you sure?
[root@fubar ~]# rpm -qf /usr/sbin/smbd samba-3.0.33-3.7.el5
Disk naming seems slightly different--we went with more standard partitions rather than LVM, and rather than /dev/sda and so on, it was /dev---arrgh, I'm not at work, and my mind just went blank--errm, xda and xdb? At any rate, it will be fairly obvious.
Nope, all comes as sda/sdb etc. if you don't use LVM unless your external storage is funky.
assume (but note, that's an assumption, not knowledge), one could call and ask things like, Where can I find samba?
Erm, maybe one of these should contain it? (It is pretty visible during the installation).
[root@hobbit Oracle Unbreakable Linux]# mount -o loop -t iso9660 Enterprise-R5-U5-Server-x86_64-dvd.iso /mnt/mnt1 [root@hobbit Oracle Unbreakable Linux]# cd /mnt/mnt1 [root@hobbit Oracle Unbreakable Linux]# find /mnt/mnt1 -iname samba* /mnt/mnt1/Server/samba-3.0.33-3.28.el5.x86_64.rpm /mnt/mnt1/Server/samba-client-3.0.33-3.28.el5.x86_64.rpm /mnt/mnt1/Server/samba-common-3.0.33-3.28.el5.i386.rpm /mnt/mnt1/Server/samba-common-3.0.33-3.28.el5.x86_64.rpm /mnt/mnt1/Server/samba-swat-3.0.33-3.28.el5.x86_64.rpm /mnt/mnt1/Server/samba3x-3.3.8-0.51.el5.x86_64.rpm /mnt/mnt1/Server/samba3x-client-3.3.8-0.51.el5.x86_64.rpm /mnt/mnt1/Server/samba3x-common-3.3.8-0.51.el5.x86_64.rpm /mnt/mnt1/Server/samba3x-doc-3.3.8-0.51.el5.x86_64.rpm /mnt/mnt1/Server/samba3x-domainjoin-gui-3.3.8-0.51.el5.x86_64.rpm /mnt/mnt1/Server/samba3x-swat-3.3.8-0.51.el5.x86_64.rpm /mnt/mnt1/Server/samba3x-winbind-3.3.8-0.51.el5.i386.rpm /mnt/mnt1/Server/samba3x-winbind-devel-3.3.8-0.51.el5.i386.rpm /mnt/mnt1/Server/samba3x-winbind-devel-3.3.8-0.51.el5.x86_64.rpm /mnt/mnt1/Server/samba3x-winbind-3.3.8-0.51.el5.x86_64.rpm
(Samba client and samba common was available, but no smbd. Again, I did NO googling on this, just did yum provides */smbd, got no results, and then used WinSCP.)
It appears you didn't even look for it. :)
Hakan Koseoglu a écrit :
On 24 November 2010 11:57, Scott Robbins scottro@nyc.rr.com wrote:
Well, it's pretty similar. I'm not sure if I messed up the installation
It's almost identical. Oracle kernel parameters & dependencies can be automatically installed during installation albeit later on Oracle RDBMS installation gives you warnings since it doesn't like the default values in /etc/sysctl.conf as populated by OEL.
Thanks for the detailed answers. I'm currently downloading the set of five CDs. I already got the first ISO (my connection is slow), burnt it and installed it on a spare machine. The installer looks indeed quite the same as RHEL/CentOS (and makes you even wonder if they changed something besides the name :oD). Since I have to wait for the other ISOs to download, I went for a minimal install, e. g. everything unchecked, even [Base]. Unfortunately, contrary to CentOS, this way of doing things doesn't get me Yum, and a manual install of Yum apparently requires some packages that are not on the first CD. So I'll simply wait for the other ISOs to download and go for something less minimal.
I'm especially curious to know about the default Yum repos config : where does Oracle Linux have its repos and its updates ? Do I have to subscribe to something to get them or are they free ? Well, I guess I'll have to wait for my download to finish to get the answer.
Cheers,
Niki
I'm especially curious to know about the default Yum repos config : where does Oracle Linux have its repos and its updates ? Do I have to subscribe to something to get them or are they free ? Well, I guess I'll have to wait for my download to finish to get the answer.
The yum repo is public but just contains the releases, no intermediate update packages. Those require an account on the ULN. See too
http://www.oracle.com/us/technologies/027617.pdf
The ULN is used by up2date, not yum.
https://linux.oracle.com/uln_faq.html
Cheers,
Niki
Alexander
On Wed, Nov 24, 2010 at 12:05:39PM +0000, Hakan Koseoglu wrote:
On 24 November 2010 11:57, Scott Robbins scottro@nyc.rr.com wrote:
Well, it's pretty similar. I'm not sure if I messed up the installation I didn't find a samba package, but as this was for one oracle developer to transfer files, we just used winscp. (I also didn't look hard for a samba package, at present, this is just a test.)
Are you sure?
Yes, (I've snipped the rest, where you indicate other places I should have looked, but thank you.)
As I said somewhere, I suspect I messed up the install, as I had to later manually populate /etc/yum.repos.d after googling for where it should be.
Disk naming seems slightly different--we went with more standard partitions rather than LVM, and rather than /dev/sda and so on, it was /dev---arrgh, I'm not at work, and my mind just went blank--errm, xda and xdb? At any rate, it will be fairly obvious.
Nope, all comes as sda/sdb etc. if you don't use LVM unless your external storage is funky.
Hrrm. Ok, this is installation on a VMware esx server, using, when asked to choose guest, an oracle machine. Devices are xvda and xvdb.
(Maybe something to do with it being on ESXi? On another, ESX 3.5 server, all RH is /dev/sda).
Along with, I'm sure, the OP, I want to thank you for this post, it makes me realize that if we definitely do this for real, there are obviously some things I missed.
On 24 November 2010 15:13, Scott Robbins scottro@nyc.rr.com wrote:
Nope, all comes as sda/sdb etc. if you don't use LVM unless your external storage is funky.
Hrrm. Ok, this is installation on a VMware esx server, using, when asked to choose guest, an oracle machine. Devices are xvda and xvdb.
Aaah, that's interesting. I've done it on an ESX 4.x server but got none of those. On the other hand, I didn't say it was an Oracle machine, just a RHEL one (didn't found it necessary to distinguish one from other, we tend to deploy CentOS on our ESX server for test/dev environments but we have a couple of RHEL and OEL ones as well. That's worth trying again. :)
Along with, I'm sure, the OP, I want to thank you for this post, it makes me realize that if we definitely do this for real, there are obviously some things I missed.
OEL is a funny one. The only reason it exists is to destroy the upstream. They're completely unlike CentOS in mentality. Their main reason of existence is cutting RHEL from support revenue. Our PHBs decided to use OEL for customers since we're an Oracle shop at work so getting all licences & support from a single source makes accounting easier. In any case, after a typical Oracle Enterprise licence calculation RHEL or OEL seems like peanuts. What worries me is with OEL eating the support revenue from RHEL and simultaneously being dependent on RHEL for upstream dev & patches, it's not a long-term viable situation, it's not even a partnership.
There are other little things why we would go for OEL, one being the OCFS2 when we do shared-storage clusters. Reading the small pring gives you the impression that Oracle won't support OCFS2 unless it's OEL. I'm not sure that's true but hey, that's what's been decided at work.
As I mentioned, the other funny thing is if you choose Oracle-validated package, it stuffs your sysconf.ctl with values. Then try installing Oracle 11g (R1 or R2) on it, the installer barfs up warnings about various kernel parameters being wrong.
There's a public yum repo but as Alexander has mentioned, not much of a use.
I'm not a big fan of OEL. I'd rather use upstream with paying customers and CentOS internally. Unfortunately this decision is out of my hands.
On Wed, Nov 24, 2010 at 04:30:32PM +0000, Hakan Koseoglu wrote:
On 24 November 2010 15:13, Scott Robbins scottro@nyc.rr.com wrote:
Along with, I'm sure, the OP, I want to thank you for this post, it makes me realize that if we definitely do this for real, there are obviously some things I missed.
OEL is a funny one. The only reason it exists is to destroy the upstream. They're completely unlike CentOS in mentality. Their main reason of existence is cutting RHEL from support revenue. Our PHBs decided to use OEL for customers since we're an Oracle shop at work so getting all licences & support from a single source makes accounting easier.
In our case, it has something to do with support--we are still discussing this--my own take is that the problem won't be with the platform, so we should use CentOS, or if management wants to be sure of paid support, which does make sense, use RH. Oracle is saying there may be issues of aspect X not being supported if we don't do all of this on Oracle.
In any case, after a typical Oracle Enterprise licence
calculation RHEL or OEL seems like peanuts.
LOL. Yes, literally out loud. It just echoes some of what one our web developers said.
What worries me is with
OEL eating the support revenue from RHEL and simultaneously being dependent on RHEL for upstream dev & patches, it's not a long-term viable situation, it's not even a partnership.
There are other little things why we would go for OEL, one being the OCFS2 when we do shared-storage clusters. Reading the small pring gives you the impression that Oracle won't support OCFS2 unless it's OEL. I'm not sure that's true but hey, that's what's been decided at work.
Yeah, see above.
As I mentioned, the other funny thing is if you choose Oracle-validated package, it stuffs your sysconf.ctl with values. Then try installing Oracle 11g (R1 or R2) on it, the installer barfs up warnings about various kernel parameters being wrong.
There's a public yum repo but as Alexander has mentioned, not much of a use.
I think that's what I took (using a recommendation that I *thought* was from Oracle's site), which may explain the samba thing. I didn't choose it during installation, although, since for our uses, that WinSCP is an easier and just as effective solution, that particular thing didn't matter--still, I bet that's it.
I'm not a big fan of OEL. I'd rather use upstream with paying customers and CentOS internally. Unfortunately this decision is out of my hands.
+1. Once again, many thanks for the input.