Unbound stopped and won't start

cmb

You make note of the contents of root.key before deleting it?

kejianshi

I'm sorry - I didn't.

The nearest reason I can guess this may have happened is a write to the file interrupted by power flicker/outage.

The ups is currently needing a battery swap.

cmb

Was hoping to get a lead on the root cause there. Seems it's happened to roughly a half dozen people, but none have reported what the contents of root.key were before deleting it.

If you happen to see it again (seems unlikely), or anyone else that happens upon this thread in the future seeing it, please note the contents of the file. Diag>Command, run:

cat /var/unbound/root.key

Or download /var/unbound/root.key from same page or via scp. The contents should be text, so cat should suffice.

doktornotor

@cmb:

but none have reported what the contents of root.key were before deleting it.

You mean like this one? https://forum.pfsense.org/index.php?topic=87357.msg479617#msg479617 - there's some inetd nonsense in there.

kejianshi

CMB - I have a few of these running here and there so if it happens again I will take a look at whats inside the file before I send it to bit heaven.

doktornotor - Yep - Thats exactly the errors it was throwing out.

cmb

Thanks, I looked through all those threads and missed the contents in that one. Looks like it's ending up with contents of other files in /var/ which would indicate the file wasn't fsynced by unbound after being written out. unbound-anchor also fails to create/update root.key if has invalid contents.

Should be fixed.
https://redmine.pfsense.org/issues/5334

Also reporting upstream to unbound as it should be doing that fsync and doesn't appear to be.

kejianshi

That is cool. I like fixed things (-:

cmb

Unbound fixed the missing fsync for a future release.
https://www.nlnetlabs.nl/bugs-script/show_bug.cgi?id=712

The fsync I added should fix it in the mean time.

kejianshi

Even better…

So I guess my bad batteries were less of a curse than I thought.

Now I need to replace them... From 8k miles away...

doktornotor

@cmb:

Should be fixed.
https://redmine.pfsense.org/issues/5334

Cannot reproduce the original issue (ZFS on the test rigs doesn't seem to suffer from any of similar "features") but intentionally screwing the anchors file gets recovered just fine now…