Decomission Servers and Subcomponents

[traviskw15]

What are best practices to decomission servers along w/ their subcomponents?

From my understanding in the out-of-box configuration retiring a parent doesnt retire its subcomponents; NICs, disks, file systems, and similar child CIs remain active/installed status..and subcomponent CI's never go through a life cycle stage. This creates stale or orphaned CI's.

What are best practices for implementing a solution and keeping good CMDB health?

[drbob]

I understand why you're asking this it seems wrong that it doesn't cascade but we need to consider context...

Does anyone actually use and manage these lower level CIs ?

Could there be other processes that manage them ?

I believe that most of the time these lower CIs are irrelevant - yes, out of curiosity I may have dug down to find the S/N of a RAM module in a particular computer but what am I doing with that - am I really tracking hardware at such a precise level ?...

If I swapped memory between two systems do I need to record that ? What is the expectation (does the RAM CI change relationship to it's parent computer CI or does it get set to retired and a new record get created under the new system) ?

If I re-partition an disk what happens to the file system CIs ? This is less obvious than the RAM question as RAM is physical so could be a CI that we track but disk partitions/file systems are more of a virtual type.

I _have_ observed that "installed software" that gets uninstalled is left in the table in a retired state. That's easier to test than moving RAM/repartitioning disks 🙂 Installed software is closer to the disk partition/filesystem type than RAM as it is also not physical.

There may be some businesses where this needs to be tracked (computer manufacturers, maybe) but I would guess that 99% don't care. So, annoying as it may be we can just ignore it. The CMDB Health dashboard is configurable for the classes it includes so if these lower classes are not included then you're consistent in ignoring them in the health context - both from user and platform perspective.

So, the real question here is... Are you asking because you have a real need to track these or because you've been looking at CI Health and noticed an anomaly ?

To the wider audience - has anyone had use cases where they've had to do this ? Do we think my hypothetical "99% don't care" is reasonable ? Is there an easy way to change this (for the "1%") ?