Consider all failure modes and improve/document handling wherever possible.
Ensure all relevant internal state is easily available in support situations e.g. via lvmdump.
Documentation might include: man page, docs team guide, knowledge base.
Internal state might be dumped via signal and/or client commands.
Review & extend/reduce log messages.
Failure modes incl. daemon dying/freezing, uevents getting lost/internal state inconsistent (incl. e.g. a missed 'remove' event)
I think that this BZ needs to be broken up into actual tasks. Ideas:
- debug and non-debug logging
- on-demand state dump (for lvmdump &c.)
- documentation of failure modes out of our control (udev bugs)
- user level documentation (but I need to know what is actually missing? we
have example.conf and a manpage, maybe need to update the LVM guide?)
Yes, new bugs for each of those and set them to 'block' this bug.
Then change the Keyword on this bug from 'FutureFeature' to 'Tracking'. (The new bugs will themselves be 'FutureFeature'.)
Just an instance of the problem related to resilience/robustness: bug #823918.
the sub-BZs of this have been all POST'd now, what remains to do is documentation. I'll POST this BZ as soon as that's done. I think we are on track to move lvmetad out of tech preview for 6.4, pending QA results...
Things seem to be going well so far. I have checked in some upstream doc updates. The handbook is in Steve's hands, and I believe he has the info he needs. Therefore, I'm closing this one.