Re: Long check points

Hans Siebrand wrote: > > Hans Siebrand wrote in message ... > >Performance on one of the servers at my site is taking an exceptionally log > >time for checkpoints, in the order of 120 to 160 seconds. > > > >What should I be looking for to shorten the check point length other than > >decreasing CKPTINTVL. > > > > > >Hans > > Out put from the system is: > > OS: SCO 5.0.5 OpenServer > IO: KAIO > > CLEANERS 8 # Number of buffer cleaner processes > LRUS 12 # Number of LRU queues > LRU_MAX_DIRTY 4 # LRU percent dirty begin cleaning limit > LRU_MIN_DIRTY 2 # LRU percent dirty end cleaning limit> > cpasx001:informix:p4 /tmp >onstat -p > > Informix Dynamic Server Version 7.30.UC2 -- On-Line -- Up 14:10:28 -- > 258048 Kbytes > > Profile > dskreads pagreads bufreads %cached dskwrits pagwrits bufwrits %cached > 6937873 4082615 105137757 93.40 785323 1304339 13490224 94.18 > > isamtot open start read write rewrite delete commit > rollbk > 51664971 152396 1274885 45273034 1032036 6901 25909 15982 2 > > 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 0 105 1657.80 692.63 86 335 > > bufwaits lokwaits lockreqs deadlks dltouts ckpwaits compress seqscans > 300807 0 49283024 0 0 44 37206 5922 > > ixda-RA idx-RA da-RA RA-pgsused lchwaits > 272958 10034 4112289 4389657 18875 > > cpasx001:informix:p4 /tmp >onstat -D > > Informix Dynamic Server Version 7.30.UC2 -- On-Line -- Up 14:14:11 -- > 258048 Kbytes > > Dbspaces > address number flags fchunk nchunks flags owner name > 8c84613c 1 1 1 1 N informix rootdbs > 8c847a18 2 1 2 1 N informix physical_logs > 8c847ad4 3 1 3 1 N informix > logical_logs_1 > 8c847b90 4 1 4 1 N informix > logical_logs_2 > 8c847c4c 5 1 5 1 N informix temp_1 > 8c847d08 6 2001 6 1 N T informix temp_2 > 8c847dc4 7 1 7 1 N informix temp_3 > 8c847e80 8 2001 8 1 N T informix temp_4 > 8c847f3c 9 1 9 1 N informix temp_5 > 8c86d828 10 2001 10 1 N T informix temp_6 > 8c86d8e4 11 1 11 1 N informix temp_7 > 8c86d9a0 12 1 12 3 N informix wslis > 8c86da5c 13 1 15 3 N informix wslistest > 8c86db18 14 1 18 1 N informix datamart > 8c86dbd4 15 1 19 1 N informix gis > 8c86dc90 16 1 20 2 N informix trim > 8c86dd4c 17 1 21 1 N informix reports > 8c86de08 18 1 22 1 N informix misc > 8c86dec4 19 1 24 1 N informix trimtest > 19 active, 2047 maximum > > Chunks > address chk/dbs offset page Rd page Wr pathname > 8c8461f8 1 1 0 1329 148 /dev/rdbirootl > 8c846654 2 2 0 12 191965 /dev/rdbiphyslogl > 8c846730 3 3 0 220712 131796 /dev/rdbilogs1l > 8c84680c 4 4 0 217191 131209 /dev/rdbilogs2l > 8c8468e8 5 5 0 400 1249 /dev/rdbic1l > 8c8469c4 6 6 0 44084 37232 /dev/rdbic2l > 8c846aa0 7 7 0 849 1254 /dev/rdbic3l > 8c846b7c 8 8 0 42421 34228 /dev/rdbic4l > 8c846c58 9 9 0 5224 11411 /dev/rdbic5l > 8c846d34 10 10 0 54182 44964 /dev/rdbic6l > 8c846e10 11 11 0 833 1255 /dev/rdbic7l > 8c846eec 12 12 25600 1057580 376542 /dev/rdbic1l > 8c846fc8 13 12 25600 74212 728 /dev/rdbic2l > 8c8470a4 14 12 25600 1774178 339321 /dev/rdbic3l > 8c847180 15 13 25600 8 0 /dev/rdbic4l > 8c84725c 16 13 25600 3 0 /dev/rdbic5l > 8c847338 17 13 25600 3 0 /dev/rdbic6l > 8c847414 18 14 25600 28148 0 /dev/rdbic7l > 8c8474f0 19 15 525600 8 0 /dev/rdbic7l > 8c8475cc 20 16 575600 371851 65 /dev/rdbic7l > 8c8476a8 21 17 625600 2983 0 /dev/rdbic7l > 8c847784 22 18 675600 1198 0 /dev/rdbic7l > 8c847860 23 16 725600 192838 1040 /dev/rdbic7l > 8c84793c 24 19 825600 158 0 /dev/rdbic7l > 24 active, 2047 maximum It appears that all your updates are going to chunks 12 & 14. These chunks share disks with dbspaces temp_1 & temp_3. You could try removing these two temporary dbspaces from your DBSPACETEMP variable to see if that helps. Ultimately you need to spread those tables in chunks 12 & 14 out onto other disks to spread your IO. At the moment ALL your page cleaners are attempting to write to those disks at the same time. > onstat -F> > Informix Dynamic Server Version 7.30.UC2 -- On-Line -- Up 14:14:47 -- > 258048 Kbytes > > Fg Writes LRU Writes Chunk Writes > 105 530998 197904 How many buffers do you have? Can you add any more buffers? This would also help in the mean time while you spread that IO out. Also, how long are your LRU queues? If you already have a large number of buffers, then increase your LRUS value which might help reduce your FG writes. Cheers, -- Mark. +----------------------------------------------------------+-----------+ | Mark D. Stock mailto:mdstock@mydas.freeserve.co.uk |//////// /| | http://www.informix.com http://www.informixhandbook.com |///// / //| | http://www.iiug.org +-----------------------------------+//// / ///| | |What year 2000 bug? year 2000 bug? |/// / ////| | |year 2000 bug? year 2000 bug? year |// / /////| | |2000 bug? year 2000 bug? year 1900 |/ ////////| +----------------------+-----------------------------------+-----------+