Over the last few weeks I have switched all my CentOS 5/6 systems over to the public yum repos from Oracle. They had made it really clear that they want more community users and that they’ll consider a “switchedover” system viable for support should you need / want support.
Compared to what I knew from other RHEL-support offerings like HP, Fujitsu or RedHat themselves this is pretty end-user-friendly.
Of course they’ll be happy if you get a support contract and a ksplice license.
But most probably I’ll not be needing any support, so what else was there?
For me the reasons were:
- quicker updates with reliable release plan
I had already decided I *had* to get away from CentOS. Situations like last year where for half a year no security updates (or any other) were done, just can’t happen. RHEL wasn’t really interesting too since I don’t see myself running one system that subscribes channels and then distributes the updates and whatnot. My time isn’t free either and too many people miss that. If you assume some few $$$ for the subscription and then another 10 hrs/year for maintenance that *is* costly.
- more current supported kernel (if needed)
Where I need it I can switch to the OEL kernel that is stripped down and more current. The german iX magazine had a good review of the “UBK” vs. Stock RH Kernel vs. SLES. While RedHat had put a lot more work on the RHEL kernel, most was backporting…
- tmem support with less hassle
I can easily grab a kernel supporting TMEM to have more usable RAM
- Supported platform for running virtualized Oracle DBs
This is just a side bonus, since I run a bastardized OVM for dom0. 🙂
So what were the experiences from switching over for a small-ish setup?
I run about 15 RHEL-clone-ish virtual machines and servers and switched all of them over time.
I did it in 3 groups, deciding by usage and backup footprint mostly.
To switch I used the CentOS2OEL script from Ksplice/Oracle.
You just run the script, let it replace the yum repos and then it de-downloads the packages and replaces them. If it all works out, you reboot into the new kernel and you’re done.
I reported back all errors I found and they even gave me instructions for how to complete the migration with each error. All these errors were subsequently fixed in their new version of the script, and they also were all limited to non-standard installs I run.
All in all, that was a pretty nice experience. Most of the issues I fixed within 3-5 minutes, for one I didn’t find the solution so I used the one they suggested and it worked 🙂
Where I had stripped down dependencies, some stuff was added back by yum. That was to be expected.
Binary compatible: I saw they actually say RedHat Enterprise Linux in /etc/redhat-release. I didn’t really like that, but I surprise this is a compatibilty thing. Many times I’ve had to fiddle with that file to install certain software. But I think it is potentially worse to give a false name. Maybe I’ll ask them about it.
Mirror Speed: The mirrors are a tad slow, but the number is also increasing. It just seems demand is bigger than what they can feed it.
Infiniband: There are OFED packages and kernels, but Oracle’s own (UEK) kernels don’t ALL come with the infiniband packages. By the version history I feel they actually forgot about it. The standard RHEL-compatible kernels are as good or bad as the original ones are.
Yum updates: YES THEY ARE PROPERLY SIGNED:
(sorry for the bad screenshot, wordpress is being a ***** again)
-bash-4.1# yum –security check-update
Loaded plugins: downloadonly, fastestmirror, priorities, security
Loading mirror speeds from cached hostfile
* epel: mirror.fraunhofer.de
Limiting package lists to security relevant ones
14 package(s) needed for security, out of 20 available
dhclient.x86_64 12:4.1.1-31.P1.0.1.el6_3.1 ol6_latest
dhcp-common.x86_64 12:4.1.1-31.P1.0.1.el6_3.1 ol6_latest
glibc.x86_64 2.12-1.80.el6_3.3 ol6_latest
glibc-common.x86_64 2.12-1.80.el6_3.3 ol6_latest
kernel.x86_64 2.6.32-279.2.1.el6 ol6_latest
kernel-firmware.noarch 2.6.32-279.2.1.el6 ol6_latest
krb5-libs.x86_64 1.9-33.el6_3.2 ol6_latest
nspr.x86_64 4.9.1-2.el6_3 ol6_latest
nss.x86_64 3.13.5-1.0.1.el6_3 ol6_latest
nss-sysinit.x86_64 3.13.5-1.0.1.el6_3 ol6_latest
nss-tools.x86_64 3.13.5-1.0.1.el6_3 ol6_latest
nss-util.x86_64 3.13.5-1.el6_3 ol6_latest
openldap.x86_64 2.4.23-26.el6_3.2 ol6_latest
sudo.x86_64 1.7.4p5-13.el6_3 ol6_latest
I still haven’t switched my Cobbler servers over. That means I sometimes install CentOS just to migrate it a moment later.
I had installed Veritas Storage foundation basic and ran into some kernel module issues, the vxportal could not correctly load.
I’m wondering if that was coming down to a hardware issue in my server, because later on I found the onboard NICs caused all kinds of issues. I didn’t have a need to re-test this yet, but probably it couldn’t hurt to take a few more stabs at testing DKMS and friends in OEL.
That’s all I could think of now.
It’s been a breeze, got feedback from nice people(*) and now my systems have a lot of added stability by getting updates when I need them.
(if you consider Oracle the evil, then well, I just talked to a small part named Ksplice.
No, actually the same goes for the Xen devs, for the Oracle VM project people.
You can’t say that for some Linux distros)