Oracle's Ksplice Live Kernel Patching Picks Up Known Exploit Detection

[starshipeleven]

Look closer... It's a burned blade server.

Look more closer, it's EVE online's burned blade server. Yes I know what it is.

Just because a server has got hot-swapping components doesn't mean it's save. These components can still kill the entire server.

Not really, in this case only this thing and the neighboring ones in the rack were lost. But it's a blade server. The server themselves are hot-swappable modules in a 2-3U rack chassis, because in a blade system you dont' care that much about each individual server.

[starshipeleven]

You can get only so far with backups on the same machine.

Who said there are no backups. Redundancy is not a backup. Having a cluster is NOT a backup.

Tape drives and cold storage servers (local and remote) are backups.

Software failure will take it down, and even redundant systems can't fail. If you aren't running a cluster of redundant machines, preferably multiple clusters in multiple locations, you're asking for downtime.

You little shits keep ignoring that in real life you usually pay for stuff.

"multiple clusters in multiple locations" is VERY expensive and it's NEVER EVER done unless the actual workload demands it.

[Guest]

Tape drives and cold storage servers (local and remote) are backups.

Yes, they're offline backups, which are important as well. However, having online backups is also important, especially if you don't want downtime.

[starshipeleven]

Yes, they're offline backups, which are important as well. However, having online backups is also important, especially if you don't want downtime.

No they are just "backups". A backup implies 100% data protection.

The thing you call "online backups" is redundancy, hot spares, or high availability. RAID, clusters and so on. They ensure that failures won't disrupt the service, but will NOT provide 100% data protection as they won't protect from the most common way of data loss (and downtime because someone has to restore backups): user error.

If someone deletes data by mistake, all the RAID and the cluster and whatever will mirror the change instantly in all your system. The backup will not.

[sdack]

Not really

Not really? ... Oh yeah really!

[Guest]

You little shits keep ignoring that in real life you usually pay for stuff.

If the customer doesn't have enough money in their budget for redundancy, they obviously don't care about downtime. Keeping servers around costs money, but it should be a marginal cost of running a business, and definitely not the kind of expense where you might want to cheap out.

[skeevy420]

Why is people assuming you can freely and easily allocate dozens of servers for everything? In many cases you have only one. The customer won't pay for more than one.

I'm sorry, but if they don't have a 2nd test system or a good server running multiple VMs that allows for both testing and production environments then they're not doing it right. Even a random fucktard such as myself knows that.

Plus, they way I figure it is if one isn't willing or is too cheap to pay for that, they damn sure aren't going to pay a Linux kernel dev to write live kernel patches for the times it doesn't do it's automagic stuff.

[starshipeleven]

Not really? ... Oh yeah really!

If the entire server is an hot-swappable board on a backplane this should hint at the fact that it's not really for critical applications, and they expect them to fail relatively often and be replaced.

So yeah, "not really".

[starshipeleven]

If the customer doesn't have enough money in their budget for redundancy, they obviously don't care about downtime.

This is your own unwashed savage claim. Real life is different.

They still want as little downtime as possible even if they don't pay for a replicated setup.

[starshipeleven]

I'm sorry, but if they don't have a 2nd test system or a good server running multiple VMs that allows for both testing and production environments then they're not doing it right.

You can run a "test system" in a VM on a workstation. In general "test systems" are not really in production (or at least should not be) nor powerful enough for production use.

tl:dr your test system is NOT redundancy.

Plus, they way I figure it is if one isn't willing or is too cheap to pay for that, they damn sure aren't going to pay a Linux kernel dev to write live kernel patches for the times it doesn't do it's automagic stuff.

Does not mean they don't want it though.

The server OS vendors are doing it, to cater to these type of customers so they can have a competitive advantage by offering "kernel security patching without downtime" since to the contrary of that DoMiNeLa10 thinks everyone will like to limit downtime on their server even if they don't want to buy massive clusters replicated overseas.

Who is usually making these patches again? RedHat, SUSE, Ubuntu and Oracle, that BY TOTAL COINCIDENCE make most of their revenue with servers.

That's all an Illuminati conspiracy, I tell you.