BTSA1 IndexInput EOF error

Hi All, Strange one here. Informix 12.1.FC4 on Solaris 10. I have 5 different instances with various forms of a database utilizing a BTS index. Nothing is different across the instances with the index or the table in question. A programmer is trying to do an INSERT into the table with the BTS index, and is receiving the following error **in only 1 instance**. The other 4 instances work perfectly fine. ----------- INSERT INTO name_dtl (nm_id, nm_typ, first_nm, middle_nm, last_nm, full_nm, org_nm, srch_first_nm, srch_middle_nm, srch_last_nm, srch_full_nm, srch_org_nm, ce_id, add_dtm, add_by, mod_dtm, mod_by, pgm_nm) SELECT 0, nm_typ, first_nm, middle_nm, last_nm, full_nm, org_nm, srch_first_nm, srch_middle_nm, srch_last_nm, srch_full_nm, srch_org_nm, 163361, add_dtm, add_by, mod_dtm, mod_by, 't3124' FROM name_dtl WHERE nm_id = 89094; (BTSA1) - bts clucene error: IO error: IndexInput read past EOF --------------- The full_nm column (varchar 150) has the bts index on it. Running the Select returns the appropriate record. I did find this thread, but it does not address our specific issue -- we have no problems with space and there are no errors in the message log. http://www.iiug.org/forums/ids/index.cgi/read/24381 Any ideas? What else would you need to help diagnose the issue? Thanks, Mike Hoffman

Have you tried running something like oncheck -cdI against it? (I haven't used BTS, so I don't know if it works...) --EEM > -----Original Message----- > From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of > MICHAEL HOFFMAN > Sent: Monday, March 14, 2016 17:54 PM > To: ids@iiug.org > Subject: BTSA1 IndexInput EOF error [36739] > > Hi All, > Strange one here. Informix 12.1.FC4 on Solaris 10. > > I have 5 different instances with various forms of a database utilizing > a BTS index. Nothing is different across the instances with the index > or the table in question. > > A programmer is trying to do an INSERT into the table with the BTS > index, and is receiving the following error **in only 1 instance**. The > other 4 instances work perfectly fine. > ----------- > INSERT INTO name_dtl > (nm_id, nm_typ, first_nm, middle_nm, last_nm, full_nm, org_nm, > srch_first_nm, srch_middle_nm, srch_last_nm, srch_full_nm, srch_org_nm, > ce_id, add_dtm, add_by, mod_dtm, mod_by, pgm_nm) SELECT 0, nm_typ, > first_nm, middle_nm, last_nm, full_nm, org_nm, srch_first_nm, > srch_middle_nm, srch_last_nm, srch_full_nm, srch_org_nm, 163361, > add_dtm, add_by, mod_dtm, mod_by, 't3124' > FROM name_dtl > WHERE nm_id = 89094;> > (BTSA1) - bts clucene error: IO error: IndexInput read past EOF > --------------- > > The full_nm column (varchar 150) has the bts index on it. > > Running the Select returns the appropriate record. > > I did find this thread, but it does not address our specific issue -- > we have no problems with space and there are no errors in the message > log. > http://www.iiug.org/forums/ids/index.cgi/read/24381 > > Any ideas? What else would you need to help diagnose the issue? > > Thanks, > Mike Hoffman > > > *********************************************************************** > ******** > Forum Note: Use "Reply" to post a response in the discussion forum.

Since the bts index is considered an External index, oncheck passes successfully, but doesn't seem to really do any verfications. > Have you tried running something like oncheck -cdI against it? (I haven't > used BTS, so I don't know if it works...) > --EEM

Hi Michael, The BTSA1 error is generated when the CLucene Open Source storage=20 interface detects an IO problem. the addition message: IndexInput read=20 past EOF +is generated "short" read from an index file (for example it=20 expects to read 16 bytes and gets something less than 16 bytes). It was=20 a seen in the first releases of BTS (11.10/11.50) but I have not seen it=20 reported in 11.70 or 12.10 defect.. Have you seen any additional messages = in the online log at the time of the error like a no space message on a=20 sbspace or tempsbspace.=20 The storage interface is used for both the persistent storage of the=20 index, the temp storage (in temp sbspace/sbspace or in memory) used to=20 modify the index. If the error is happening while reading the persistent=20 storage, then you could see it on queries too. I don't have any work around to this problem since its an error you should = never see. I can only offer some generic solutions to try to resolve the=20 issue: Compact the index: begin work; execute function bts=5Findex=5Fcompact('your index name');=20 commit on success, rollback on failure. Essentially bts=5Findex=5Fcompact() will only compact the index if there we= re=20 deletions done since the last compact operation. If you do not delete=20 rows from the table, the compact operation is a no-op. The last option to try is to drop and recreate the bts index. ----- If you can generate a repro of the problem, I would open an PMR get=20 support to open a product defect so we can investigate why CLucene is=20 generating the error. -- Mark. Mark Ashworth=20 Office phone: +1 (905) 413-5033=20 Email: mailto:ashworth@ca.ibm.com=20 Check out my blog=20 From: "MICHAEL HOFFMAN" <offdisc@gmail.com> To: ids@iiug.org Date: 03/14/2016 08:18 PM Subject: BTSA1 IndexInput EOF error [36739] Sent by: ids-bounces@iiug.org Hi All,=20 Strange one here. Informix 12.1.FC4 on Solaris 10.=20 I have 5 different instances with various forms of a database utilizing a=20 BTS=20 index. Nothing is different across the instances with the index or the=20 table=20 in question.=20 A programmer is trying to do an INSERT into the table with the BTS index,=20 and=20 is receiving the following error **in only 1 instance**. The other 4=20 instances=20 work perfectly fine.=20 -----------=20 INSERT INTO name=5Fdtl=20 (nm=5Fid, nm=5Ftyp, first=5Fnm, middle=5Fnm, last=5Fnm, full=5Fnm, org=5Fnm= ,=20srch=5Ffirst=5Fnm, srch=5Fmiddle=5Fnm, srch=5Flast=5Fnm, srch=5Ffull=5Fnm, = srch=5Forg=5Fnm,=20 ce=5Fid, add=5Fdtm, add=5Fby, mod=5Fdtm, mod=5Fby, pgm=5Fnm)=20 SELECT 0, nm=5Ftyp, first=5Fnm, middle=5Fnm, last=5Fnm, full=5Fnm, org=5Fnm= ,=20srch=5Ffirst=5Fnm, srch=5Fmiddle=5Fnm, srch=5Flast=5Fnm, srch=5Ffull=5Fnm, = srch=5Forg=5Fnm,=20 163361, add=5Fdtm, add=5Fby, mod=5Fdtm, mod=5Fby, 't3124'=20 FROM name=5Fdtl=20 WHERE nm=5Fid =3D 89094;=20 (BTSA1) - bts clucene error: IO error: IndexInput read past EOF=20 ---------------=20 The full=5Fnm column (varchar 150) has the bts index on it.=20 Running the Select returns the appropriate record.=20 I did find this thread, but it does not address our specific issue -- we=20 have=20 no problems with space and there are no errors in the message log.=20 http://www.iiug.org/forums/ids/index.cgi/read/24381=20 Any ideas? What else would you need to help diagnose the issue?=20 Thanks,=20 Mike Hoffman=20 ***************************************************************************= ****=20 Forum Note: Use "Reply" to post a response in the discussion forum.=20

Is it too big to make a rebuild impractical? I know that's kind of a Micro$oft type solution, but it may work. > -----Original Message----- > From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of > MICHAEL HOFFMAN > Sent: Tuesday, March 15, 2016 12:28 PM > To: ids@iiug.org > Subject: Re: RE: BTSA1 IndexInput EOF error [36749] > > Since the bts index is considered an External index, oncheck passes > successfully, but doesn't seem to really do any verfications. > > > Have you tried running something like oncheck -cdI against it? (I > > haven't used BTS, so I don't know if it works...) --EEM > > > *********************************************************************** > ******** > Forum Note: Use "Reply" to post a response in the discussion forum.