EAServer and ASA11 -- connection cache question

[forums.sybase.com]

Our web application has a web server and database server. Both use Windows
2003. The web server has EAS 5.1 and components written in PB9.0.3. The db
server uses ASA 11.0.1. The two communicate with one another via a
connection case in EAServer.

Sometimes it is necessary to drop a connection in Sybase Central on the
database server. For example, we recently had a problem with an index which
I tried to rebuild. However, a table lock prevented me from doing so.
Killing the connection allowed me to rebuild the index which worked fine.
However, I've discovered that when dropping a connection in Sybase Central,
our web application begins to act very erratically and the following error
appears repeatedly in the Jag log:

cm__odbc_error: SQLError: -308, SQLState: HY000, [Sybase][ODBC Driver][SQL
Anywhere]Connection was terminated.

When this happens, it is necessary to restart the Jaguar service to get the
web app working normally again. I've been able to replicate the problem on
our development server even when the index is ok so it seems that the
connection cache is really sensitive to disconnects. Has anyonw else
experienced this problem? Is there a setting/option in either the
connection cache or EAS that would prevent this from happening?

Thanks, Tom

[Alex Whitney]

Do you have the 'Enable connection sanity check' option enabled for the
cache? This will cause the cache to test the connection before reusing it.

Alex

"forums.sybase.com" <trol...@microedge.com> wrote in message
news:4b21428e@forums-1-dub...

[Libby]

Yes, this option is checked. Tom

"Alex Whitney" <ale...@pdq.net> wrote in message
news:4b218159$1@forums-1-dub...

[forums.sybase.com]

I tried turning sanity check off but ran into the same problem with our web
application -- pages not rendering properly, items disappearing, fonts and
colors changing, etc. Our web app has a unique look and feel for each
client that uses it and the settings that control this branding are all
stored in the database.

The one thing that was different is that with sanity check off, it didn't
check for the dropped connection and therefore an odbc error did not show up
in the jaguar log.

Tom

"Alex Whitney" <ale...@pdq.net> wrote in message
news:4b218159$1@forums-1-dub...

[Alex Whitney]

what options do you have set in the dbparm string when you connect to the
database? Do you have UseContextObject set?

Alex

"Libby" <trol...@yahoo.com> wrote in message
news:4b21dc26$1@forums-1-dub...

[Tom Rolseth]

We have the following in our dbparm string:
"UseContextObject='Yes',CacheName='xyz'", nothing else

Doesn't sanity check try to re-establish the connection if it detects one
has failed? It doesn't seem to be handling this correctly based on what
I've read. Maybe this is a bug in EAS 5.2?

Tom

"Alex Whitney" <ale...@pdq.net> wrote in message

news:4b22aabf$1@forums-1-dub...

[Mark Maslow]

Do you have stateful components that connect to the DB? If the
component disconnects from the cache before deactivating, the next time
it needs a connection, sanity check should handle reconnecting. But if
the component never disconnects from the cache, sanity checking isn't
going to be of any help.

In article <4b21428e@forums-1-dub>, trol...@microedge.com says...

[Alex Whitney]

In our environment, dropping a connection does not cause any issues. We are
on EAS 5.1 buld 51016

Our connect string looks like yours.

what is your sanity check statement set to?

from the props file for your connection cache:
com.sybase.jaguar.conncache.check=select 1 from dummy

Alex

"Tom Rolseth" <trol...@microedge.com> wrote in message
news:4b22b7ab$1@forums-1-dub...