On 2018-10-29 15:42, Aleksej Lebedev wrote:
On 2018-10-29 15:12, Tim Darby wrote:
This is an issue I've reported:
http://bugs.dragonflybsd.org/issues/2915
My way around this is to create the new master and use cpdup to copy
the PFS slave to it.
Tim
Hi, Tim.
Thanks for your reply.
Your problem does look similar to mine. But I now think it might not
be a bug: latest transactions might not be commited.
Reading through the manual once more I noticed there is a command
"hammer synctid" which syncs the PFS and gives you the TID that you
might use to check that mirror-copy/stream reached it.
I will experiment with it a bit more and let you know if the problem
resolves.
I ended up deleting my PFS restored from a backup and re-restoring it
again, now making sure its sync-end-tid is the same as in the backup
PFS.
After it's done I checked that it contained exactly the same data - it
did.
Then I tried to do mirror-copy back, i.e. from just restored PFS to the
backup - it worked instantly.
Then I upgraded my restored PFS to master, made some changes and checked
if mirror-copy/stream picks them up - all worked as expected.
So it looks like the only problem was last time that I upgraded my
recovered PFS too early - before hammer committed the last transaction.
Having said that, I tried to reproduce the scenario from the ticket you
reported and indeed there was a problem. I even checked that the
sync-end-tid is the same after mirror-copy - still the problem
persisted.
I am surprised this ticket is 2 years old and there are no comments. If
it's really a bug in hammer, it very important.
--
Aleksej Lebedev
