Troubleshooting
Problem
DB2 will get trap when running SELECT with WHERE IN clause on table with a few thousands partitions and EXPLAIN MODE set
Symptom
The conditions required to hit this issue are:
- A SELECT statement over a table with a huge number of partitions (around 1000 or more).
- In the SELECT a WHERE IN clause with enough values to have around 500 partitions eliminated or more.
- Before running the SELECT a 'db2 set current explain mode explain' should have been run.
The stack will be corrupted (overwritten) but the rawstack dump will show ascii numbers similar to this:
0x0A000001B47F01D0 : 3832 382D 3833 332C 2038 3336 2D38 3337 828-833, 836-837
0x0A000001B47F01E0 : 2C20 3834 312C 2038 3434 2D38 3435 2C20 , 841, 844-845,
0x0A000001B47F01F0 : 3834 392D 3835 302C 2038 3532 2C20 3835 849-850, 852, 85
0x0A000001B47F0200 : 392D 3836 302C 2038 3639 2C20 3839 322C 9-860, 869, 892,
0x0A000001B47F0210 : 2038 3934 2C20 3839 392C 2039 3037 2D39 894, 899, 907-9
The stack trace should look like this:
0x09000000280ADA58 @137@sqlnx_exp_datapart_info__FP23sqlnx_exp_lolepop_parmsPUiPUcT2PPcPi + 0x890
Stack traceback unavailable.
Resolving The Problem
1.The workaround is to avoid running with the 'explain mode' on.
2. We have APAR IT18969 opened for this trap issue , it will fix in v10.5fp9 , you may request a fix on top of your db2level .
Was this topic helpful?
Document Information
Modified date:
16 June 2018
UID
swg21999245