[BlueOnyx:27436] Re: 5210R/5211R YUM updates (CodeIgniter update)

[Michael Stauber]

Hi Dirk,

> I really like easy migrate, but when I switch from CentOS7 to
> AL 8 or 9, I have to go through all the sites because the
> directory structure of the webroot has changed. That is
> time-consuming. But unavoidable.

Yeah, the directory structure being the same between EL8/9 and EL10 will 
make this possible as far as that goes.

But like said: It's the OS and the binaries that I'm concerned about. 
I'm sure if AlmaLinux again publishes an "elevate" script, it'll take 
care of most of it. I could augment that with a plugin that handles the 
BlueOnyx side of things as far as the RPMs out of our official BlueOnyx 
YUM repositories go.

But it gets iffy for installed PKGs with binary code from the shop. 
Which would typically mean at least AV-SPAM, PHP, Easy-Backup and 
OpenVPN. These contain binaries compiled for the specific OS depending 
on the platform where they are installed.

So even if the "elevate" works for a "naked" BlueOnyx without these 
PKGs? It will fail hard for those with these PKGs installed and I'd need 
to find a way to script around this.

And as far as that goes: This requires an amount of testing and trial 
and error that it isn't really funny. We're talking about a 2-3 weeks 
all out daily effort to make it solid. And lots and lots of testing.

I can test this easily with virtualized servers and if it breaks, I just 
blow the target VPS away and start over. But if such an error happens 
with a real physical server on the client side later on? That box is 
then in a sorry state and that client is in distress.

And that's the part where we need to talk "business", too: While this 
for sure would be a "nice to have feature"? If the procedure goes wrong, 
the client will be displeased. As this most likely was an "in situ" 
upgrade it means a production server is now in a sorry state and may 
require a restore from a backup, or expert intervention. That's not a 
favorable user experience and will displease the client and his clients. 
People also will get unhappy with me, because I delivered something that 
didn't work for them.

Likewise: The three weeks all out effort to work out a viable procedure 
and to test it covering a multitude of scenarios? That's a lot of work. 
I can see the gain it has for clients and end-users, but is it 
marketable? Can I sell this as a service and make it worth the time I 
have to spend on it? Even then: If it's a paid service, it has to work 
100% of the time ALL the time. And that's something I *may* not be able 
to guarantee with a comfortable safety margin. At least I can't say 
until I've had a chance to play with "elevate" for EL9 -> EL10 and have 
a more solid idea on any technical obstacles that may arise.

I'm not opposed to doing it and giving it a try. But that can only 
happen when BlueOnyx 5212R is finished *and* AlmaLinux has an "elevate" 
on offer that helps us with the EL9 -> EL10 part.

Then I'll give it an honest effort and will check how viable it might be 
to get this working in a way that we can offer it.

-- 
With best regards

Michael Stauber