[clue-tech] Custom partition scheme?

[Nate Duehr]

> In my environment, mostly virtualized servers in VMware Server on Linux, LVM snapshots make simple backups possible. Without LVM snapshots, backups would be a nightmare.
>

An LVM snapshot isn't a copy of the data... it just marks where the 
snapshot took place, and then knows how to "get back there"... a Hansel 
and Gretel trail of breadcrumbs so to speak.  (It keeps track of the 
original inode numbers.)

I have used snapshots on LVM and other filesystem types that are capable 
of snapshots to "freeze" the filesystem so a relatively slower 
backup-to-tape system could then back that snapshot up over a weekend's 
time, so that no changes are taking place to the data, while the backup 
is running.  That works well.

But if LVM itself fails, the snapshots aren't copies, and they're gone 
gone gone... they're not a substitute for getting the data off, and onto 
another platter, tape, CD, whatever...

As a tool to "freeze" everything so a backup can be taken, tools like 
LVM are great.  (Don't forget to "quiet" your databases and other things 
on that disk, if you have things there that aren't just files, but are 
actively being written to like binary DB spaces, etc.)

My point was only that the limitations of the tool must be known, and 
one of those is that LVM *can* fail, just like anything else...

It doesn't happen often, but I've seen it (once)... it took out the 
machine.  I've also seen professional grade but similar tools like 
Veritas barf and take out all of someone's data too... there's all sorts 
of ugly once-in-a-lifetime failures that the only thing that will fix 
them is good backups, and a bare-metal recovery plan!

So if you meant that LVM is a useful tool if you plan to use it as a way 
to "quiet" your filesystem you're backing up to somewhere else, I 
definitely agree, and building with LVM in mind up-front is required. 
  But having LVM on by default for folks without a real plan on how to 
manage their systems, is kinda silly on the part of many Linux distros, 
and just adds complexity they may not (or may...) need.

I don't ascribe to the "turn it all on, we might need it" philosophy of 
system management, because every 100 lines of code has at least one bug 
in it... the less overall code running, the less overall risk of 
failures... in my humble opinion.  There's enough complexity in the 
computing world without turning on things like LVM by default long 
before a new sysadmin even knows what it is, and what they can (and 
can't) use it for.  :-)

Nate