philipp.reisner at linbit
May 23, 2011, 9:49 AM
Post #1 of 1
This is a bugfix only release in of the 8.3 code branch.
The most subtlety bug fixed with this release is in the
area of the max_bio_size negotiation:
Over the course of the last releases we where driving up
the max_bio_size supported by DRBD. We had an optimistic
negotiation of the agreed max_bio_size in place.
Now, if one created a device, made it primary, connected
it to a peer node with data, and attached the local disk
after all that, he could cause DRBD to shrink the
max_bio_size with that attach operation.
But shrinking the max_bio_size is a problem if a BIO gets
created with the old (larger) max_bio_size and gets
submitted when the new (smaller) max_bio_size is set.
The other fixes are quite on the same page:
Rather not so common cases.
Interesting for long distance replication users, is the bugfix
for establishing connections.
Please note: This is a release candidate, not intended for production.
Please help with testing!
* Fixed handling of read errors during online verify runs
* Fixes for possible distributed deadlocks during online verify and
checksum based resync processes
* Fixed handling of IO errors with an on-io-error=pass-on policy; the
disk state goes to inconsistent
* Fixed the behaviour in case ping-timeout and ping-int are set to the
* Fix for connecting on high latency network links
* Fixed state transitions if fence-peer handler returns after connection was
* Made the max_bio_size negotiation conservative instead of optimistic
* Network timeouts for mirrored write requests are now always enforced,
no longer only when the socket is congested
: Dipl-Ing Philipp Reisner
: LINBIT | Your Way to High Availability
: Tel: +43-1-8178292-50, Fax: +43-1-8178292-82
DRBD(R) and LINBIT(R) are registered trademarks of LINBIT, Austria.
drbd-announce mailing list
drbd-announce [at] lists