Last week 8.1.7 . 4. After upgrading to 10.2.0.1, there were many problems. One day after the successful run, the AP staff reported that tnsping was very slow and sometimes it would take over 20 thousand Hao seconds, most of the time more than 10 thousand. It wasn't long before the Production Line reflected that the barcode flushing speed was slow. Check alter log and find the following error:
Errors in file D: \ IBM \ dump \ udump \ ibm_ora_2788.trc:
ORA-07445: exception encountered: core dump [access_violation] [kcbgtcr + 4479] [PC: 0x706507] [ADDR: 0 xffffffffffffffff] [unable_to_read] []
D: \ IBM \ dump \ udump \ ibm_ora_2788.trc the following error occurs:
Opiino: attach failed! Error =-1 ifvp = 0000000000000000
I searched many posts online and did not find any answers. I posted two posts on itpub and there is no answer.
Ask for help from a friend who found similar problems on Metalink and gave me the URL. It turns out this is 10.2.0 A bug in. 1 has been fixed in a later version. It is 10.2.0.1 that the user occupies a large amount of memory in the connection pool during connection, probably 3 m Many, and this bug causes the memory to be released normally. It is easier to solve the problem.
First, in c: \ oracle \ product \ 10.2.0 Add the following parameters to \ db_1 \ Network \ admin \ sqlnet. ora:
Sqlnet. inbound_connect_timeout = 0
This can reduce the connection pool memory used by each user 2 m
Second, cancel the sga_max_size parameter to reduce the SGA 1.4 GB (Total 6g Memory, the original SGA gave 2.5 GB )
Third, restart Oracle
After the database restarts OK, it is found that the CPU usage is reduced, and tnsping is restored to the normal 20 Hao seconds. Let the production line try and reflect that it is back to normal.
It worked normally for a week, 10.2.0 The. 3 patch has also been down. It is planned to take time to update the patch. The AP staff said it was hard to get it done. Don't try again. It's almost dead. The manager hopes to upgrade the patch. Since it is an Oracle bug, there will inevitably be no more problems. I am a perfect person. I also hope to upgrade the patch. But on the one hand, the AP staff opposed it and I didn't have time to test the patch. The production line worked overtime over the weekend and could not be stopped, so I didn't upgrade it.
On the last Sunday evening, the AP staff called home to repeat the problem. Reconnect to the company and find it is an old problem. Restart the Oracle service and ask the AP staff to try again. the problem persists. Restart the computer (OS) and restore to normal. It seems that the memory cannot be released after oracle is restarted. On the other hand, the patch must be upgraded.
Upgrading the database has brought a lot of problems, but I feel very sad that it has brought a lot of trouble to the manager. Today, I met a senior DBA who said that upgrading the Oracle version would usually have a hard time for a quarter. We went through about two weeks. He said it was actually quite smooth. After hearing this, with a little relief