Re: onstat -F output

This reinforces the idea from yesterday, that you do not have enough pages updated between checkpoints to force LRU writes, given your current settings. Note that during the measured period, you have only 66,897 page writes. Owing to the imprecise amount of time ('almost an hour'), let's say this was over a period of 50 minutes. With a checkpoint interval of 300 seconds, we would have 10 checkpoints, assuming that the log file was not filling up and forcing more frequent checkpoints. 66,897 page writes / 10 checkpoints = 6690 pages per checkpoint / 8 LRU queues = 836 dirty pages per LRU queue at checkpoint time. This is barely over half of the number of dirty pages necessary to force an LRU write. 32000 buffers / 8 LRU queues = 4000 pages per LRU queue * .4 (LRUMAXDIRTY / 100) = 1600 dirty pages to force an LRU write. If the measured period was actually 60 minutes, we have less than 700 dirty pages per LRU queue, using the same calculations. Note that this includes some large assumptions - update activity is evenly distributed across the measured period, updates are distributed across the LRU queues evenly, etc. Even if there were a little fluctuation in the above, it still probably would not reach the threshold of 1600 dirty pages, so there would be no LRU writes. Try lowering your LRUMAXDIRTY and LRUMINDIRTY and you will start seeing LRU writes. You may even determine that the values should be 1 and 0 (as Art suggested). By the way, how long are your checkpoints taking? Mark Collins mcollins@us.dhl.com > > On Fri, 12 Feb 1999 richard_kofler@my-dejanews.com wrote: > > > In article <79v8ge$d5j$1@news.xmission.com>, > > Joseph Cullipher <joseph@cannonexpress.com> wrote: > > > > > > Hi all, > > > I have Informix 7.30.UC2 running on SCO 5.04 and have noticed this on > > > my onstat -F output > > > > > > Fg Writes LRU Writes Chunk Writes > > > 0 0 18463 > > > > > > address flusher state data > > > 874484c4 0 I 0 = 0X0 > > > 87448974 1 I 0 = 0X0 > > > 87448e24 2 I 0 = 0X0 > > > states: Exit Idle Chunk Lru > > > > > > This is an OLTP system so the LRU writes should be something other than 0 > > > am I correct? Also is noticed that even when a check point occurs my flushers > > > stay idle. Am I missing something here? > > > > > > here is my onconfig file > > > > > > > > > > Art and Mark are right > > It might pay off to look at the ratio buffer reads vs. buffer writes and > > your read- and write cache rate. Maybe you are running heavily read bound > > and if you do not have many writes (= delete,update,insert) > > at all you maybe will never see any LRU Writes > > As to my understanding, onstat -F gives cumulative counts for LRU and > > chunk writes. How long have these accumulated to read the value shown in > > your posting (i.e. what was your uptime then??) > > So maybe let us see the results of this action: > > Onstat -z (i.e. reset of all counters) > > After one hour of peak activity do > > onstat -Fp and post that and I'm sure that I will learn> > something from Art's answer to that posting ... > > > > Regards > > Dick > > > > Richard Kofler > > debis Systemhaus EDVg > > Vienna / Austria > > > > -----------== Posted via Deja News, The Discussion Network ==---------- > > http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own > > > OK here you are. I did the onstat -z then waited almost a hour and this is > what I got. > > Informix Dynamic Server Version 7.30.UC2 -- On-Line -- Up 02:44:44 -- 102400 Kbytes > > > Fg Writes LRU Writes Chunk Writes > 0 0 3717 > > address flusher state data > 874484c4 0 I 0 = 0X0 > 87448974 1 I 0 = 0X0 > 87448e24 2 I 0 = 0X0 > states: Exit Idle Chunk Lru > > Profile > dskreads pagreads bufreads %cached dskwrits pagwrits bufwrits %cached > 40413 96535 2812961 98.56 20372 66897 126398 83.88 > > isamtot open start read write rewrite delete commit rollbk > 2174071 28119 50262 1888480 31977 1935 692 1566 0 > > gp_read gp_write gp_rewrt gp_del gp_alloc gp_free gp_curs > 0 0 0 0 0 0 0 > > ovlock ovuserthread ovbuff usercpu syscpu numckpts flushes > 0 0 0 350.70 55.42 33 66 > > bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans > 1895 1 4385074 0 0 1 311 1716 > > ixda-RA idx-RA da-RA RA-pgsused lchwaits > 12402 2372 7121 21370 1807 > > Does this shed some light on what is happening? > > Joe >