-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256 Bertrand,
On 8/19/14, 2:23 PM, Bertrand Renuart wrote: > > > On 19/08/14 16:26, "Christopher Schultz" > <ch...@christopherschultz.net> wrote: >> >> On 8/18/14, 1:05 PM, Wes Clark wrote: >>> One of the primary conditions for which we want to mark the >>> connection for death is when a rollback fails with an >>> exception. An example of this kind of error is a network >>> error. >> >> Won't this happen automatically? If the network connection dies, >> the server should auto-rollback the transaction. The JDBC >> connection will fail due to the same reason and the pool will >> re-establish a new Connection. Are you sure you have to manage >> this yourself? > > For the pool to re-establish a new connection you have to set > testOnBorrow=3Dtrue, otherwise you have to wait until the next time > the PoolCleaner kicks in (provided you have testOnIdle=true). Wes¹ > proposition speeds up the detection of the dead connection without > requiring a validation on borrow. Aah, I see. That wasn't entirely clear from the original posting, but it makes sense given what Wes was saying. >>> We do this with: >>> >>> PooledConnection pooledConnection = >>> conn.unwrap(PooledConnection.class); >>> pooledConnection.setDiscarded(true); >>> >>> I think I'll modify my code so that if an exception is thrown, >>> instead of unconditionally evicting the connection from the >>> pool, run the validation query right then and only evict if it >>> fails. >> >> This won't catch intermittent network errors, which seems like >> the only really valid use case for you. > > We also faced some weird jdbc drivers that forced us to drop a > connection after some specific exceptions have occurred. If not, we > got strange behaviour with some queries but not all (and of course > the validation query was ok :-( Here also we followed an approach > similar to Wes=B9 proposal. I'm interested in which JDBC driver that affected. Can you share? > PooledConnection pooledConnection = > conn.unwrap(PooledConnection.class); > pooledConnection.setDiscarded(true); I wonder if calling Connection.abort (from the public java.sql.Connection API) would be a bit safer. I haven't checked the tomcat-pool code to see if this is a no-op or if it works as desired. If it doesn't work, it might be nice to support that kind of thing for these cases. - -chris -----BEGIN PGP SIGNATURE----- Version: GnuPG v1 Comment: GPGTools - http://gpgtools.org iQIcBAEBCAAGBQJT87GOAAoJEBzwKT+lPKRY1OYP/2W6XEODiKg/t7kn83leDPUh miOqBTJ84rv5BQP5m9HZ/a37xWmitDsQxrCMWRd885pc/dga9tfOiSjdFJsg7Z2u rOJqiP5ItPMaTYtpDxblDV3vehC5qkSE6U4ZbiSfDpq35pgDGqTUUy5iOQqWMHA5 yQGb8IbM6uKfGAjj06iIHeRvy9l2iiyFL/DFy+XeYWttrfA3GK5h5e012DYMH0Rm aTaDtxjoKs98ktt4oHKEdRMH8caU87rjNA6RkIjCThpXcI+gYv8pZ+Xnujot3WWO JQkbU7lJzMS/Al/QzhUjDA+hDQz/1CUYSc9UHHohRZHp4ghO98h/rKgCADwWDFta V6SHNmnTcnNCUlp+cF6KVhAB4IvKdApNaxmx7stjLmLPtYCCYhLDBuUOiVdwOlZ2 nAtsch1Otd6RMkEFF4yw9HKNdbuvYqqcP9IFlwO4kUfQOubD0hl4JjJiSK20DXAi gyWKLXBSu9jHcZLMiADTPEFQrA+aB+mWL5zn+lNIVik0F7vJodidP2Gtbew1Wa6/ 3FPb56X/xzMa4IdtmirHktZrz64Ky5Gf86IbWs1r1MS3W7nhpsRCgLZYHkmpUB+S rNIHwq+2/VpUNQeAfj8AGCQw720fb539a6vsQoYiQT5euPf/zT4T8XwFvheInuE/ eAQyeZzOcJhY4w70d23H =6Isa -----END PGP SIGNATURE----- --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org
