When we went to Linux servers, we also went with JDBC instead of ODBC for connections, because I wasn't that familiar with setting up ODBC in a Linux environment. Witango doesn't seem to like fixing broken JDBC connections as easily as it does ODBC connections. I eventually had time to get comfortable setting up ODBC in Linux, and have moved back to ODBC because of the problems mentioned in this thread.
/John On Mon, Oct 13, 2008 at 5:35 PM, Shannon Henderson <[EMAIL PROTECTED]>wrote: > > All of the solutions, except for resetting the datasourcelife, require some > server side monitoring. > > I haven't done it, but in theory this should work. If you check for > <@error number1=113> and set a value somewhere, you could run a cronjob to > check for the value and restart the server if necessary. You might be able > to monitor your logs for the same error and restart the server then. > We use a cron that executes a witango app that has a single db call > (something like select 'x' from dual) and returns an error if it fails. We > decided not to be specific about the error. If the app gets any error, we > restart the server. Of course the downside to this is that it looks like > your witango server is acting up if your database server is unavailable for > some reason. We found that most of our production db connectivity errors > were this one so it hasn't really been an issue. > > I wish I could be more help, but frankly we've pretty much given up making > this better. We've got our duct tape and bailing wire solution that will > hopefully hold out until we can replace it with something more stable. > > -sh > > On Oct 13, 2008, at 1:18 PM, MC Tay wrote: > > Thank you Shannon for your tips. How do I trap 113 error and restart the > Witango programmatically? Is it on the Witango side? > > When we have both the app and db on the same server, it works fine. You are > right, the problem with the remote db connection could due to firewall, drop > packet etc. > > MC > > > At 09:36 AM 10/13/2008, you wrote: > > It's a known issue, and it is not specific to Oracle. We've seen the same > error frequently for Filemaker JDBC and Oracle connections for a couple of > years now. The best solution anyone has come up with is to trap the error > and then issue a restart to the server process. There does not appear to be > an easy way to kill off the specific connection. > > It appears that the error occurs when the database connection is not > properly closed, usually due to some networking interference (dropped > packets, firewall killing idle connections, etc.) We were able to reduce the > number of failures by eliminating some of the potential points of networking > failure between servers A and B. If your database server is remote, this may > not be an option for you. The other thing we did was make sure the > DATASOURCELIFE was set to be shorter than likely network timeouts. This > created a separate problem in Oracle 9+ where idle sessions were left on the > Oracle side, eventually using up all of the available users and > significantly annoying our dba, so we ended up scheduling witango server > restarts anyway to clear those. > > Depending on your setup, namely the number of connections and the amount of > site traffic you're talking about, you can try trapping the 113 error and > then setting the DATASOURCELIFE to 0, immediately hitting the same db > connection again (to timeout the datasource). Then you reset the > DATASOURCELIFE to it's previous setting. That should force the connection to > be retried the third time most of the time. This works pretty well in our > dev environment, but it turned out to be impractical with the sheer number > of connections we work with in production. > > A couple of people have contacted me on and off list saying they've had the > same issues, so I'm sure a few of us would be happy to hear of a better > solution should you come across one. > > --sh > > > > On Oct 13, 2008, at 12:55 AM, MC Tay wrote: > > Hi: > > I have encountering lost database connection and need some help. > > I have a Witango application (Server A) accessing Oracle database (Server > B) on a remote site. It works fine not until may be 2-3 hours later the > database connection is lost. I have to restart Witango service on Server A > and it works again. But, few hours later the database connection is lost > again. > > Any idea how to fix this and is it a known problem? > > Thanks! > > MC > > ________________________________________________________________________ > TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf > > > > ________________________________________________________________________ > TO UNSUBSCRIBE: Go to > http://www.witango.com/developer/maillist.taf > > ________________________________________________________________________ > TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf > > > ________________________________________________________________________ > TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf > > -- /John ________________________________________________________________________ TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf
