bmurray at snf.stanford.edu
Mon Mar 4 01:11:28 PST 2002
Remote users have been able to access Coral until I restarted the servers
a few minutes ago. However, local users lost access to coral around 12:36
this afternoon. I've spent the last few hours looking at the database
tables and logs before I restarted the servers.
Without going into all the details, I believe that the admin manager's
JDBC connection to the database hung at about 12:30pm today. Since
each manager has a single JDBC connection to the database, this ultimately
results in the failure of the manager. I need to find a way to confirm
my hypothesis. Then I can evaluate the two possible solutions. We
could use a connection pool which presents some tricky technical issues
in our architecture. Or we could consider options for periodically
opening a new connection to the database or attempting to detect
connection failures and opening a new connection.
I'll sleep on this one and talk to you in the morning.
More information about the coral