[Bucardo-general] Performance issues... (4.5.x)
michelle at sorbs.net
Wed May 1 12:07:48 UTC 2013
Michelle Sullivan wrote:
> Greg Sabino Mullane wrote:
>> On Tue, Apr 30, 2013 at 07:49:11PM +0200, Michelle Sullivan wrote:
>>> I'm having trouble here... without bucardo (4.x) I get a 10x speed
>>> Please suggestions on queries and expectations...
>> With the triggers enabled, during the 10s delay, run:
>> SELECT relation::regclass, * FROM pg_locks where not granted;
>> Any rows that show up are strong clues for a locking issue.
> This is a pretty consistent view it's all 'pg_listener' with very
> occasionally the htimestamps table appearing:
Ummm... yeah that fixed it....
sorbs=# vacuum full analyse verbose pg_catalog.pg_listener;
INFO: vacuuming "pg_catalog.pg_listener"
INFO: "pg_listener": found 32 removable, 437 nonremovable row versions
in 32306 pages
DETAIL: 430 dead row versions cannot be removed yet.
Nonremovable row versions range from 96 to 96 bytes long.
There were 2616316 unused item pointers.
Total free space (including removable row versions) is 253366428 bytes.
32296 pages are or will become empty, including 1663 at the end of the
30643 pages containing 240321728 free bytes are potential move destinations.
CPU 0.45s/0.05u sec elapsed 0.51 sec.
INFO: "pg_listener": moved 437 row versions, truncated 32306 to 6 pages
DETAIL: CPU 0.41s/0.26u sec elapsed 13.53 sec.
INFO: analyzing "pg_catalog.pg_listener"
INFO: "pg_listener": scanned 6 of 6 pages, containing 7 live rows and
438 dead rows; 7 rows in sample, 7 estimated total rows
More information about the Bucardo-general