Re: Insert Query issue

Look what's happening to the cell servers (sysstat across all 4 nodes) (node, stat, bad value, good value)
1 cell physical IO bytes sent directly to DB node to balance CPU 1,156,614,749,830 0
2 cell physical IO bytes sent directly to DB node to balance CPU 69,966,913,016 0
4 cell physical IO bytes sent directly to DB node to balance CPU 248,957,618,776 0

1 cell num bytes in passthru during predicate offload 430,937,849,856 0 2 cell num bytes in passthru during predicate offload 3,434,520,576 0 4 cell num bytes in passthru during predicate offload 3,299,098,624 0

1 cell physical IO interconnect bytes returned by smart scan 6,366,091,608,830 803,494,139,600
2 cell physical IO interconnect bytes returned by smart scan 122,454,213,640 136,610,305,640
4 cell physical IO interconnect bytes returned by smart scan 256,740,327,790 231,764,927,070

1 physical read requests optimized 26,803,215 4,551,705 2 physical read requests optimized 1,155,409 3,080,777 3 physical read requests optimized 187 31 4 physical read requests optimized 4,161,123 1,269,049

1 physical read total multi block requests 25,167,050 4,392,758 2 physical read total multi block requests 1,148,400 2,992,359 3 physical read total multi block requests 154 17 4 physical read total multi block requests 3,854,283 529,528

1 physical reads direct 2,774,110,979 478,369,098

2 physical reads direct 131,330,451 363,692,790
3 physical reads direct 0 1,260
4 physical reads direct 481,489,849 54,190,916

1 table scan blocks gotten 1,785,345,451 322,123,204 2 table scan blocks gotten 28,199,043 36,062,181 3 table scan blocks gotten 494,457 248,725 4 table scan blocks gotten 473,425,109 130,520,953 During the bad period node 1 is kicking the storage devices to death (with some help from node 4); so much so that the cell servers are sending compressed CUs to the database servers so that they have to do the decompression.
Disks and flash cache are also under far more pressure from node 1 (and 4), so all read requests are slower.

Regards
Jonathan Lewis

--
http://www.freelists.org/webpage/oracle-l