Please don't reply to lustre-devel. Instead, comment in Bugzilla by using the 
following link:
https://bugzilla.lustre.org/show_bug.cgi?id=4778



Created an attachment (id=9228)
Please don't reply to lustre-devel. Instead, comment in Bugzilla by using the 
following link:
 --> (https://bugzilla.lustre.org/attachment.cgi?id=9228&action=view)
minor remaining cleanups for inspection

Before I was testing this bug I was having repeated problems with precreate
hitting LBUGs due to existing objects, and I thought I was hitting this bug. 
It
turned out to be a bad interaction with one of the regression test
failure-injection hooks and the async object creates from bug 4900 which allow
testing to continue while one OST is still doing object recovery.

Attached is a small patch that disables the OBD_FAIL_OST_ENOENT hook if we are
doing orphan recovery to avoid skipping objects during orphan cleanup and then
tripping the "Serious error: object already exists" assertion later.

Also, in filter_handle_precreate() it skips orphan destroy requests from "old"
connections, similar to what is in the other branch of the "if" in the previous

patch, since it doesn't make sense to handle either of them if there was a
reconnection and the request has been resent.

It also clarifies some of the OSC-side debug messages to avoid confusing the
"last used" and "next to use" values.

_______________________________________________
Lustre-devel mailing list
[email protected]
https://mail.clusterfs.com/mailman/listinfo/lustre-devel

Reply via email to