On Feb 28, 2013, at 8:37 AM, Ilias Bertsimas <email address hidden> wrote:
> Yes it is flow control pause that stalls the cluster. That can be solved with innodb_blocking_buffer_pool_restore=ON
Ok, great. This bit is "not a bug" technically.
> But the side effect I mentioned is the restoration of the buffer pool pages twice.
>
> Once when wsrep tries to recover uuid:gtid in
> WSREP: Running position recovery with --log_error=/tmp/tmp.nci5tZ9xZF
> And the second time where it should recover the buffer pool pages if the node was not part of a galera cluster.
Yes, precisely. The wsrep_recover should not do an LRU reload. It's not technically a bug, but it's superfluous and obviously extra delay bringing up the node.
On Feb 28, 2013, at 8:37 AM, Ilias Bertsimas <email address hidden> wrote:
> Yes it is flow control pause that stalls the cluster. That can be solved with innodb_ blocking_ buffer_ pool_restore= ON
Ok, great. This bit is "not a bug" technically.
> But the side effect I mentioned is the restoration of the buffer pool pages twice. /tmp/tmp. nci5tZ9xZF
>
> Once when wsrep tries to recover uuid:gtid in
> WSREP: Running position recovery with --log_error=
> And the second time where it should recover the buffer pool pages if the node was not part of a galera cluster.
Yes, precisely. The wsrep_recover should not do an LRU reload. It's not technically a bug, but it's superfluous and obviously extra delay bringing up the node.
Jay Janssen, MySQL Consulting Lead, Percona about.me/ jay.janssen www.percona. com/live/ mysql-conferenc e-2013/
http://
Percona Live in Santa Clara, CA April 22nd-25th 2013
http://