Re: Dirty LRU exceeding the LRU_MAX of 2%

More information to this problem. RAID5 is already being used before I started here please don't flame me for using raid5. Please send me the information about the use of raid5 so I can educate my sys admins. They think it is great. I have searched the iiug cdi but can not find the doc about raid5. I think the problem with my LRU is CPU contention. vmstat reports idle at 0 wa at 90+ (probably because of raid5). I think it is just taking a long time for the CPUs to service the flushers because they are busy with the disk I/O. Does this sound locial. Your thoughts would be appreciated. Thanks "ART KAGEL, BLOOMBERG/ NEW YORK" <KAGEL@bloomberg.net> on 11/11/99 08:06:23 AM To: cc: (bcc: Kernoal Stephens/STS/AUTOZONE) Subject: Re: Dirty LRU exceeding the LRU_MAX of 2% No the buffer pool is independent of the transaction mechanism. Those buffers should be flushing if enough resources are available. I'd take a shared memory dump and some onstats and call tech support. Something is not right in Denmark! ---- Original Msg from: Kernoal Stephens <Kernoal.Stephens@AUTOZONE.COM> At: 11/11 8:09 KAIO is on. I think the application is doing a lot of work and not commiting. Would this hold the buffers until checkpoint or until commit? "Art S. Kagel" <kagel@bloomberg.net> on 11/10/99 05:13:07 PM Please respond to kagel@bloomberg.net To: Kernoal Stephens/STS/AUTOZONE cc: Subject: Re: Dirty LRU exceeding the LRU_MAX of 2% If you are not running with KAIO enabled then the problem is that you only have 2 AIO VPs. Art S. Kagel Kernoal.Stephens@AUTOZONE.COM wrote: > > Here is what I am seeing. In onstat -R the dirty q's are over 2% dirty. > onstat -F shows all the flushers as Idle.> > I know I have seen this in the post but cannot find it in search. > > Attached. > onstat -R output > onstat -F output> onconfig output > IDS 7.30.UC6 > AIX 4.3.2 > 4CPU > 512MB > > Informix Dynamic Server Version 7.30.UC6 -- On-Line -- Up 00:28:33 -- > 141568 Kbytes > > 24 buffer LRU queue pairs priority levels > # f/m pair total % of length LOW MED_LOW MED_HIGH HIGH > 0 f 825 96.8% 799 1 304 494 0 > 1 m 3.2% 26 0 8 18 0 > 2 f 821 99.0% 813 0 344 469 0 > 3 M 1.0% 8 0 0 8 0 > 4 f 832 95.0% 790 0 294 496 0 > 5 m 5.0% 42 0 7 35 0 > 6 f 832 96.4% 802 0 271 531 0 > 7 m 3.6% 30 0 11 19 0 > 8 f 829 99.0% 821 0 272 549 0 > 9 M 1.0% 8 0 0 8 0 > 10 f 834 95.3% 795 1 318 476 0 > 11 m 4.7% 39 0 13 26 0 > 12 f 837 97.3% 814 0 283 531 0 > 13 m 2.7% 23 0 7 16 0 > 14 f 843 96.4% 813 0 343 470 0 > 15 m 3.6% 30 0 11 19 0 > 16 f 825 95.4% 787 0 313 474 0 > 17 m 4.6% 38 0 18 20 0 > 18 f 820 95.1% 780 0 299 481 0 > 19 m 4.9% 40 0 13 27 0 > 20 f 827 94.7% 783 0 282 501 0 > 21 m 5.3% 44 0 10 34 0 > 22 F 826 98.7% 815 0 330 485 0 > 23 M 1.3% 11 0 2 9 0 > 24 f 838 97.4% 816 0 368 448 0 > 25 m 2.6% 22 0 4 18 0 > 26 f 818 97.1% 794 0 329 465 0 > 27 m 2.9% 24 0 7 17 0 > 28 f 832 98.7% 821 0 307 514 0 > 29 m 1.3% 11 0 2 9 0 > 30 f 832 97.5% 811 1 367 443 0 > 31 m 2.5% 21 0 7 14 0 > 32 f 836 99.0% 828 0 312 516 0 > 33 M 1.0% 8 0 0 8 0 > 34 f 827 98.9% 818 0 335 483 0 > 35 m 1.1% 9 0 1 8 0 > 36 f 824 99.0% 816 1 300 515 0 > 37 M 1.0% 8 0 0 8 0 > 38 f 839 95.5% 801 0 318 483 0 > 39 m 4.5% 38 0 12 26 0 > 40 f 844 97.6% 824 0 356 468 0 > 41 m 2.4% 20 0 9 11 0 > 42 f 809 98.8% 799 0 314 485 0 > 43 M 1.2% 10 0 0 10 0 > 44 f 825 97.8% 807 0 312 495 0 > 45 m 2.2% 18 0 4 14 0 > 46 f 815 99.0% 807 0 342 465 0 > 47 M 1.0% 8 0 0 8 0 > 536 dirty, 19890 queued, 20000 total, 32768 hash buckets, 4096 buffer size > start clean at 2% (of pair total) dirty, or 16 buffs dirty, stop at 1% > > onstat -F > Informix Dynamic Server Version 7.30.UC6 -- On-Line -- Up 00:28:38 --> 141568 Kbytes > > Fg Writes LRU Writes Chunk Writes > 0 25320 1572 > > address flusher state data > 400404c8 0 I 0 = 0X0 > 4004097c 1 I 0 = 0X0 > 40040e30 2 I 0 = 0X0 > 400412e4 3 I 0 = 0X0 > 40041798 4 I 0 = 0X0 > 40041c4c 5 I 0 = 0X0 > 40042100 6 I 0 = 0X0 > 400425b4 7 I 0 = 0X0 > 40042a68 8 I 0 = 0X0 > 40042f1c 9 I 0 = 0X0 > 400433d0 10 I 0 = 0X0 > 40043884 11 I 0 = 0X0 > 40043d38 12 I 0 = 0X0 > 400441ec 13 I 0 = 0X0 > 400446a0 14 I 0 = 0X0 > 40044b54 15 I 0 = 0X0 > 40045008 16 I 0 = 0X0 > 400454bc 17 I 0 = 0X0 > 40045970 18 I 0 = 0X0 > 40045e24 19 I 0 = 0X0 > 400462d8 20 I 0 = 0X0 > 4004678c 21 I 0 = 0X0 > 40046c40 22 I 0 = 0X0 > 400470f4 23 I 0 = 0X0 > states: Exit Idle Chunk Lru > > onconfig > Informix Dynam

> Please send me the information about the use of raid5 so I can educate my > sys admins. They think it is great. Ask Art. Hi _really_ loves this question :). > I have searched the iiug cdi but can not find the doc about raid5. > > I think the problem with my LRU is CPU contention. > vmstat reports idle at 0 wa at 90+ (probably because of raid5). And what sar saing about that ? What kind of RAID you have - hardware or program ? If your CPUs are busy, of course, there is no time for cleaners to perform their job. ------------------------------------------------- With best regards, Yuri Dovgart SAP R/3, Informix technical consultant, Informix Certified Professional, Senior System Consultant System Architecture and High Availability Systems, 'Telecominvest' company Email y_dovgart@tci.ukrtel.net ICQ 39284285