RE: waiting for log buffer??

Posted in 2003

Topics: Storage & Space Management, Logging & Checkpoints, Versions, Editions & End-of-Life

Rajasekaran, Rajesh — Fri, 25 Jul 2003 07:51:51 -0500 — source: Usenet: comp.databases.informix

This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. ------_=_NextPart_001_01C352AB.84A7CAA0 Content-Type: text/plain; charset="iso-8859-1" You can try extending the size of LOGBUFF to say 64 KB ofcourse not exceeding your LOGSIZE. Also, it could be due to PHYSBUFF as well. Do you see physical log forcing checkpoint in your system ? Do y have checkpoints running apart from the CKPTINVL ? In that case,try increasing PHYSBUFF to 1024/2048 again not exceeding PHYSIZE. Please let me know the results in case if you decide to try the above things. Rajesh Rajasekaran Informix Database Administrator Forest Pharmaceuticals Inc. (314) 493-7073 rrajasekaran@forestpharm.com -----Original Message----- From: james [mailto:hohoho@erewhon.com] Sent: Thursday, July 24, 2003 9:30 PM To: informix-list@iiug.org Subject: waiting for log buffer?? I don't understand what is happening (not unusual) in a particular informix instance. ids 7.30.UC7, hpux 10.20 a data loading script is building an index on a temporary table; there are two temp dbspaces, the temp table was created 'with nolog', and the build is using multiple threads. at some point, other sessions, doing inserts and updates to unrelated tables, get a 'G' in the onstat -u first flag position; at this time, the index build session seems to be stuck in a critical section - it shows a 'X' in flag four for a period of time (minutes, not seconds) and the server is *not* in a ckp wait state The sessions stuck on 'G', waiting for a log buffer, do nothing until the index build completes, then continue merrily. The log buffers are 32k default. The instance is logging to /dev/nul. There are 8 log files of 500 MB each. KAIO is being used, 4 cpuvps If anyone could enlighten me as to what is happening, I would appreciate it. A cure would be warmly received as well. tia James ------_=_NextPart_001_01C352AB.84A7CAA0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN"> <HTML> <HEAD> <META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; = charset=3Diso-8859-1"> <META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version = 5.5.2653.12"> <TITLE>RE: waiting for log buffer??</TITLE> </HEAD> <BODY> You can try extending the size of LOGBUFF to say 64 = KB ofcourse not exceeding your LOGSIZE. Also, it could be due to PHYSBUFF as well. Do you see = physical log forcing checkpoint in your system ? Do y have checkpoints = running apart from the CKPTINVL ? In that case,try increasing PHYSBUFF to 1024/2048 = again not exceeding PHYSIZE. Please let me know the results in case if you decide = to try the above things. Rajesh Rajasekaran Informix Database Administrator Forest Pharmaceuticals Inc. (314) 493-7073 rrajasekaran@forestpharm.com -----Original Message----- From: james [<A = HREF=3D"mailto:hohoho@erewhon.com">mailto:hohoho@erewhon.com</A>]= Sent: Thursday, July 24, 2003 9:30 PM To: informix-list@iiug.org Subject: waiting for log buffer?? I don't understand what is happening (not unusual) in = a particular informix instance. ids 7.30.UC7, hpux 10.20 a data loading script is building an index on a = temporary table; there are two temp dbspaces, the temp table was created = 'with nolog', and the build is using multiple threads. at some point, other sessions, doing inserts and = updates to unrelated tables, get a 'G' in the onstat -u first flag = position; at this time, the index build session seems to be stuck in a = critical section - it shows a 'X' in flag four for a period of time = (minutes, not seconds) and the server is *not* in a ckp wait state The sessions stuck on 'G', waiting for a log buffer, = do nothing until the index build completes, then continue = merrily. The log buffers are 32k default. The instance is = logging to /dev/nul. There are 8 log files of 500 MB each. KAIO is being used, 4 cpuvps If anyone could enlighten me as to what is happening, = I would appreciate it. A cure would be warmly received as = well. tia James </BODY> </HTML> ------_=_NextPart_001_01C352AB.84A7CAA0-- sending to informix-list

↪ replying to Rajasekaran, Rajesh

james — Fri, 25 Jul 2003 20:56:42 GMT — source: Usenet: comp.databases.informix

Thank you for your interest. The physical log buffer is 128k, this the physical log is large enough that a 75% full phys log *never* triggers a ckpt, it's always the ckpt interval of 10 minutes. It was silly of me to not mention it, but yes the database is using buffered logging. I've included the onstat -l for further information. Regards, James Informix Dynamic Server Version 7.30.UC7 -- On-Line -- Up 28 days\\par > 18:56:11 -- 729528 Kbytes\\par > \\par > Physical Logging\\par > Buffer bufused bufsize numpages numwrits pages/io\\par > P-1 0 128 10267068 92652 110.81\\par > phybegin physize phypos phyused %used \\par > 3100035 50000 41209 0 0.00 \\par > \\par > Logical Logging\\par > Buffer bufused bufsize numrecs numpages numwrits recs/pages pages/io\\par > L-1 0 32 128757935 6326502 660639 20.4 9.6 \\par > \\tab Subsystem numrecs Log Space used\\par > \\tab OLDRSAM 128757935 3507979608 \\par > \\par > address number flags uniqid begin size used %used\\par > c2132d24 1 U---C-L 79303 310c385 50000 48871 97.74\\par > c2132d40 2 U-B---- 79296 31186d5 50000 50000 100.00\\par > c2132d5c 3 U-B---- 79297 3124a25 50000 50000 100.00\\par > c2132d78 4 U-B---- 79298 3130d75 50000 50000 100.00\\par > c2132d94 5 U-B---- 79299 313d0c5 50000 50000 100.00\\par > c2132db0 6 U-B---- 79300 3149415 50000 50000 100.00\\par > c2132dcc 7 U-B---- 79301 3155765 50000 50000 100.00\\par > c2132de8 8 U-B---- 79302 3161ab5 50000 50000 100.00\\par > \\par On Fri, 25 Jul 2003 07:51:51 -0500, "Rajasekaran, Rajesh" <RRajasekaran@forestpharm.com> wrote: > >This message is in MIME format. Since your mail reader does not understand >this format, some or all of this message may not be legible. > >------_=_NextPart_001_01C352AB.84A7CAA0 >Content-Type: text/plain; > charset="iso-8859-1" > >You can try extending the size of LOGBUFF to say 64 KB ofcourse not >exceeding your LOGSIZE. > >Also, it could be due to PHYSBUFF as well. Do you see physical log forcing >checkpoint in your system ? Do y have checkpoints running apart from the >CKPTINVL ? >In that case,try increasing PHYSBUFF to 1024/2048 again not exceeding >PHYSIZE. > >Please let me know the results in case if you decide to try the above >things. > >Rajesh Rajasekaran >Informix Database Administrator >Forest Pharmaceuticals Inc. >(314) 493-7073 >rrajasekaran@forestpharm.com > > >-----Original Message----- >From: james [mailto:hohoho@erewhon.com] >Sent: Thursday, July 24, 2003 9:30 PM >To: informix-list@iiug.org >Subject: waiting for log buffer?? > > >I don't understand what is happening (not unusual) in a particular >informix instance. > >ids 7.30.UC7, hpux 10.20 > >a data loading script is building an index on a temporary table; there >are two temp dbspaces, the temp table was created 'with nolog', and >the build is using multiple threads. > >at some point, other sessions, doing inserts and updates to unrelated >tables, get a 'G' in the onstat -u first flag position; at this time, >the index build session seems to be stuck in a critical section - it >shows a 'X' in flag four for a period of time (minutes, not seconds) >and the server is *not* in a ckp wait state > >The sessions stuck on 'G', waiting for a log buffer, do nothing until >the index build completes, then continue merrily. > >The log buffers are 32k default. The instance is logging to /dev/nul. >There are 8 log files of 500 MB each. > >KAIO is being used, 4 cpuvps > >If anyone could enlighten me as to what is happening, I would >appreciate it. A cure would be warmly received as well. > >tia > >James > >------_=_NextPart_001_01C352AB.84A7CAA0 >Content-Type: text/html; > charset="iso-8859-1" >Content-Transfer-Encoding: quoted-printable > ><!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN"> ><HTML> ><HEAD> ><META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; = >charset=3Diso-8859-1"> ><META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version = >5.5.2653.12"> ><TITLE>RE: waiting for log buffer??</TITLE> ></HEAD> ><BODY> > >You can try extending the size of LOGBUFF to say 64 = >KB ofcourse not exceeding your LOGSIZE. > > >Also, it could be due to PHYSBUFF as well. Do you see = >physical log forcing checkpoint in your system ? Do y have checkpoints = >running apart from the CKPTINVL ? > >In that case,try increasing PHYSBUFF to 1024/2048 = >again not exceeding PHYSIZE. > > >Please let me know the results in case if you decide = >to try the above things. > > >Rajesh Rajasekaran > Informix Database Administrator > Forest Pharmaceuticals Inc. > (314) 493-7073 > rrajasekaran@forestpharm.com > > > >-----Original Message----- > From: james [<A = >HREF=3D"mailto:hohoho@erewhon.com">mailto:hohoho@erewhon.com</A>]= > > Sent: Thursday, July 24, 2003 9:30 PM > To: informix-list@iiug.org > Subject: waiting for log buffer?? > > > >I don't understand what is happening (not unusual) in = >a particular > informix instance. > > >ids 7.30.UC7, hpux 10.20 > > >a data loading script is building an index on a = >temporary table; there > are two temp dbspaces, the temp table was created = >'with nolog', and > the build is using multiple threads. > > >at some point, other sessions, doing inserts and = >updates to unrelated > tables, get a 'G' in the onstat -u first flag = >position; at this time, > the index build session seems to be stuck in a = >critical section - it > shows a 'X' in flag four for a period of time = >(minutes, not seconds) > and the server is *not* in a ckp wait state > > >The sessions stuck on 'G', waiting for a log buffer, = >do nothing until > the index build completes, then continue = >merrily. > > >The log buffers are 32k default. The instance is = >logging to /dev/nul. > There are 8 log files of 500 MB each. > > >KAIO is being used, 4 cpuvps > > ><P

RE: waiting for log buffer??

Related threads

Glossary

Your privacy choices