Recent Features
| |||||||||||||
Commitment to Open (Red Hat News)Commitment to Open (Red Hat News)Posted Mar 4, 2011 21:06 UTC (Fri) by foom (subscriber, #14868)In reply to: Commitment to Open (Red Hat News) by ESRI Parent article: Commitment to Open (Red Hat News)
Hm. Didn't oracle make a big deal about having a kernel that was a completely new not based on RHEL's kernel? Does this change actually even affect Oracle at all?
(Log in to post comments)
Commitment to Open (Red Hat News) Posted Mar 4, 2011 21:19 UTC (Fri) by ESRI (guest, #52806) [Link]
Hmm, yes they did. You'd think Oracle would have the resources to roll their own kernels.
Or maybe they used a Fedora kernel :)
They used vanialla kernel Posted Mar 4, 2011 21:38 UTC (Fri) by khim (subscriber, #9252) [Link] Actually they explained what they did. They are rolling out their own kernel, but use userspace from RHEL - and of course they cherry-picked few changes needed to do that. Essentially they want to fix the "Solaris problem" by piggybacking on RHEL: Solaris had great kernel but poor userspace so it was really painfull to use so Oracle think they can avoid this problem by concentrating on kernel and enterprise (where the money are) and leaving userspace (where the support costs are) to RedHat. RedHat does not want this. It basically says: no problem, roll out your own kernel, if you wish, but then, please, support your userspace too.
They used vanialla kernel Posted Mar 4, 2011 21:51 UTC (Fri) by foom (subscriber, #14868) [Link]
> RedHat does not want this. It basically says: no problem, roll out your own kernel, if you wish, but then, please, support your userspace too.
But wait...RedHat is going to do that by making it hard for Oracle to steal RedHat's kernel changes? Which Oracle wasn't doing anyways...Errrrrrr, what?
PS: what's up with Oracle's kernel, anyways? Their git tree doesn't seem to have been updated in ages:
They used vanialla kernel Posted Mar 4, 2011 22:23 UTC (Fri) by jengelh (subscriber, #33263) [Link]
>PS: what's up with Oracle's kernel, anyways? Their git tree doesn't seem to have been updated in ages:
Are you surprised? Ever since the acquisition thing, everything has practically come to a halt. Last entry of onnv (below) is just about the same timeframe unbreakable.git stopped being updated.
changeset: 13149:b23a4dab3d50
They used vanialla kernel Posted Mar 4, 2011 22:11 UTC (Fri) by quotemstr (subscriber, #45331) [Link]
So why not just buy Nexenta? If you're right, it's much closer to what they're trying to do.
They used vanialla kernel Posted Mar 4, 2011 23:23 UTC (Fri) by ESRI (guest, #52806) [Link]
This is what we're doing. Costs for running NexentaStor (not Nexenta) on non-Oracle hardware are compareable to what Oracle charges now.
The risk is whether or not Nexenta and the Illumos folks have the resources to support development of drivers for newer hardware going forward... and whether or not they can count on Oracle doing Solaris source code dumps periodically to either re-base or cherry pick...
Commitment to Open (Red Hat News) Posted Mar 4, 2011 21:21 UTC (Fri) by mrshiny (subscriber, #4266) [Link]
If Oracle is customizing the vanilla kernel, then this means they can't easily cherry-pick the patches they like from RedHat's kernel.
Commitment to Open (Red Hat News) Posted Mar 5, 2011 16:13 UTC (Sat) by msnitzer (subscriber, #57232) [Link]
Sure, but even if Oracle didn't modify their 2.6.32-stable based uek kernel they'd have a hard time cherry-picking RHEL's kernel patches.
Oracle aside, various other distro kernels are based on 2.6.32-stable too. RHEL6's kernel is _not_ based on 2.6.32-stable. That said, RHEL6's kernel does contain many of the same upstream linux-2.6's "stable@kernel.org" fixes that feed into 2.6.32-stable.
But due to the various 2.6.3[45678] backports that the RHEL6.x kernel has seen the code has diverged considerably from 2.6.32-stable. This divergence is to the point that most fixes made to RHEL6's kernel are either irrelevant to or do not apply cleanly to 2.6.32-stable. The fixes would be more suitable to apply to 2.6.36-stable, 2.6.37-stable, etc. Red Hat kernel developers do mark appropriate upstream linux-2.6 fixes with: "Cc: stable@kernel.org".
So Red Hat's fixes really do flow back to "stable@kernel.org" kernels -- unfortunately (for Oracle, Novell, Ubuntu, Debian, etc) those changes are not overly useful to 2.6.32-stable based distribution kernels. NOTE: all "stable@kernel.org" kernels do not allow new features whereas the RHEL6's kernel does.
It would require quite some surgery to backport the RHEL6.x kernel changes to 2.6.32-stable (because the implicit requisite feature surgery is _not_ possible in a "stable@kernel.org" kernel). The challenge associated with trying to backport RHEL6's changes to 2.6.32-stable will only grow over time.
Oh, yeah, it does. Posted Mar 4, 2011 21:27 UTC (Fri) by khim (subscriber, #9252) [Link] Oracle selectively pulls changes needed to make RHEL programs happy, then add it's own patches and packages the result. Now they will have a dilemma: either use pure RHEL source (and forget about "latest and greatest" claims) or start actually making its own distro.
Oh, yeah, it does. Posted Mar 4, 2011 21:58 UTC (Fri) by foom (subscriber, #14868) [Link]
I'm having a hard time imagining that there's very much from RHEL5's 2.6.18 kernel which is even applicable to Oracle's 2.6.32-based kernel.
But I guess RedHat is concerned that now that they're *also* using the 2.6.32 kernel, Oracle might stop bothering to maintain their own kernel from scratch and just use RHEL's with some extra patches on top, or something.
Oh, yeah, it does. Posted Mar 4, 2011 23:31 UTC (Fri) by ESRI (guest, #52806) [Link]
My question is -- does this mean Oracle Unbreakable Linux is gaining traction? Or is this a preemptive move from RH?
I never understood why, from a customer perspective, I'd want to use OEL instead of the "real thing" (RHEL). Plus, as big as Oracle is, RH has the bigger mindshare and expertise as far as kernel hackers and engineers (at least devoted to Linux) -- not to say Oracle doesn't have some quality people, just not as many.
I never thought OEL would really take off or be a threat to RHEL ...
Oh, yeah, it does. Posted Mar 7, 2011 13:40 UTC (Mon) by michel (subscriber, #10186) [Link]
You would think that, except that Oracle can simply declare their own kernel the only one 'tested and supported' for their database. As a customer of their database, why wouldn't I use it?
Having said all that, I have no clue how Oracle is going to continue to support their kernel once they've undermined RH enough. I'm sure not many of the RH engineers want to work for oracle (apparently not many of the Sun engineers did either...). I have this impression that for Larry Ellison it's all about winning, regardless of whether he's a winner of a wasteland, rather than a large participant in a vibrant ecosystem.
Oh, yeah, it does. Posted Mar 4, 2011 23:34 UTC (Fri) by SEJeff (subscriber, #51588) [Link]
This entire discussion is about RHEL6, not RHEL5.
|
|||||||||||||