Re: Host of Errors During Large Upload

Gene Hubert wrote:
>
> Hi,
>
> I'm running Oracle 7.2 under NT 3.51. I'm currently in test mode but
> am working up to the real thing. I have two instances running with
> 64M memory on the server.
>
> My db is in ARCHIVELOG mode with automatic archiving enabled. I
> have 8, 200K log files set up with only 1 member in each log group.
>
> During a large upload of @150 Meg., I got a bunch of warnings/errors
> in my alert file.
>
> The primary Oracle messages were 00255 and 00286 which are
> something like "error during archiving" and "no members
> available or none with valid data".
>
> I would also occasionally get these lines added to the alert file:
> Thread 1 cannot allocate new log, sequence xxx
>
> Checkpoint not complete
>
> Archiver process freed from errors. No longer stopped.
>
> ARCH: Archival stopped, error occurred. Will continue retrying
>
> During the upload, the disk I was writing to nearly ran out of
> space and would have if I had not deleted some files.
>
> I'm completely baffled. I assume these are serious problems and
> must be resolved. I could not see any clear pattern to the
> various errors and warning that were written to the alert file.
>
> Any help would be greatly appreciated.
>
> gene hubert

Well here`s a first thought:

The error "Checkpoint not complete" comes from the LGWR who wants to do a log switch back to the first log group file and is waiting for your checkpoint process to terminate writing all dirty buffers to database files either because:
1) DBWR must write quite a lot of dirty buffers to files and can't finish in time. Solution: help him out with starting CKPT process if you haven't done so already.
2) your group log files members are too small. The new checkpoint process is too fast in coming back again for the next log switch and DBWR can't keep up. Solution: increase size of log group members.

Start some stats with utlbstat and utlestat and check how many log switches you have during your load. I expect that the same waiting error is happening with the ARCH process who can't start archiving before checkpoint is finished. Could it be because LGWR wants to go back to the log ARCH is waiting to archive? Check how many log groups you have. You should have at least three but you may need to increase this also for a big upload.

Hope this helps. Let me know if you succeded in correcting the problem.

Sylvie Bérubé
sberube_at_aircanada.ca Received on Mon Dec 16 1996 - 00:00:00 CST