We are facing an problem that seems to manifest itself in a general
slowdown of our entire OLTP system. During these slowdowns we have
noticed 2 things: a high number of log file synch waits and a locking
of the /proc table on UNIX (Solaris). During one such slowdown we
noticed that a delete was being executed which did not use an index.
The table being scanned during the delete is a collector table that is
populated at times with many thousands of rows which are deleted after
being consumed. The table only had a few rows in it but had many
empty blocks which were being scanned during the full table scan
caused by the delete. One slowdown corresponded perfectly with this
activity. We has since re-organized the table to eliminate the empty
blocks, but we do not understand how this may translate into high log
file synch waits or locking of the UNIX proc table. Has anyone else
hit this or do you have any knowledge to share on these symptons? If
so, could you reply to sharden (AT) uillinois (DOT) edu as well as this group?
Thank you.