Resource utilization & max session's

[newtotivoli]

Hi All,

I have a small query, One of our oracle client is using only one sessions, when it was conf to use 4 sess's even though the resources are available and that to the sess is on idlewait for the last couple of hours. Due to which schedules have been overlapping.
Idle time out is 720 on the server.
resourceutilization is on 6.
max nmp is on 10.

What might be causing this. Node is operating at 1gb and no packet loss or anything.

Please help me on this.

[sjhorvath]

To clarify- the RMAN script is allocating 4 channels, but you only see one session? Are you also saying that the session is spending a lot of time processing but not sending to TSM? I've seen this happen on large databases that have slow incremental change rates. But only when idletimeout is low like under 2 hours and RMAN is spending most of it's time looking for changed blocks. Can you post a query session output here while it is in the middle of the operation. Might want to post the RMAN script, however a DBA forum might get you better comments.

[newtotivoli]

HI Many Thanks for your reply. Yes only one sess is running at any given time during the schedule, Even though it was confg to use 4.


Here is the sess detail:

28,885 Tcp/Ip IdleW 7.8 H 469 388 Node AIX PL1PA390_ORALOG

[sjhorvath]

Idle wait for nearly 8 hours and only 300-400 bytes back and forth... If it's what I said above, seeking for changed blocks, then this would only happen on incrementals. Do the RMAN fulls work OK? In the session above, can you find out if it was a full or incr?

Questions for you:
Do other oracle backups in your shop work fine on same TSM server? (If so, it is not likely a TSM issue)
Is there anything unusual in the TSM actlog besides connecting?

For the dba:
Is the RMAN script similar to databases that work OK?
Does that database have any performance issues outside of RMAN (queries, updates and such).
Can the RMAN script be posted here?
Prove media mount success by issuing a simple backup on command line "backup current controlfile", does TSM write a backup piece to medium?

[etchingsj]

You can use the following query to monitor job progress and then contrast that to the session time.
Something like.

select process_num as Number , process, current_timestamp-start_time as Elapsed_Time ,(cast(files_processed as decimal(18,0))/cast((current_timestamp-start_time) seconds as decimal(18,0))) as Filessecond ,(cast(bytes_processed as decimal(18,0))/cast((current_timestamp-start_time) seconds as decimal(18,0)))as Bytessecond from processes

[newtotivoli]

HI sjhorvath,

Thanks for your reply.

I got it resolved, The problem was in the rman script, Somebody modified the script to use only one channel.

Thanks for your valuable advise.