Re: [sqlite] IOERR_DIR_FSYNC on main db

2008-07-15 Thread Ken
This was on an AIX 5.3 system using the JFS filesystem. I wonder what would happen on a JFS2, VFS or UFS filesystems? Ken Ken <[EMAIL PROTECTED]> wrote: Sorry but after a second read, probably don't want the O_DSYNC except for maybe an initial opening of any persistent file ( open(o_dsync), c

Re: [sqlite] IOERR_DIR_FSYNC on main db

2008-07-15 Thread Ken
Just as a side note this was on an AIX 5.3 system using JFS. I wonder what would happen if some other filesystem was in in use, such as VFS,UFS etc? Ken Ken <[EMAIL PROTECTED]> wrote: Sorry but after a second read, probably don't want the O_DSYNC except for maybe an initial opening of any pers

Re: [sqlite] IOERR_DIR_FSYNC on main db

2008-07-15 Thread Ken
Sorry but after a second read, probably don't want the O_DSYNC except for maybe an initial opening of any persistent file ( open(o_dsync), close, re-open ?). Ken <[EMAIL PROTECTED]> wrote: On AIX, you might want to open the files using O_DSYNC flag: (see) http://publibn.boulder.ibm.com/doc_link

Re: [sqlite] IOERR_DIR_FSYNC on main db

2008-07-15 Thread Ken
On AIX, you might want to open the files using O_DSYNC flag: (see) http://publibn.boulder.ibm.com/doc_link/en_US/a_doc_lib/libs/basetrf1/open.htm#HDRA1509805 Test results: Opening for rdonly kwrite(1, 0xF0373B18, 19) = 19 open("/home/ixion/T", O_RDONLY|O_LARGEFILE) = 3 f

Re: [sqlite] IOERR_DIR_FSYNC on main db

2008-07-15 Thread Ken
Dan, I'll test this with a quick bit of C code and get back to you with the truss output. Good catch Roger! ( and thanks) Thanks, Ken Dan <[EMAIL PROTECTED]> wrote: On Jul 15, 2008, at 9:00 AM, Roger Binns wrote: > -BEGIN PGP SIGNED MESSAGE- > Hash: SHA1 > > Ken wrote: > >> open("/h

Re: [sqlite] IOERR_DIR_FSYNC on main db

2008-07-14 Thread Dan
On Jul 15, 2008, at 9:00 AM, Roger Binns wrote: > -BEGIN PGP SIGNED MESSAGE- > Hash: SHA1 > > Ken wrote: > >> open("/home/ixion/ix_propagator/data/db/ajax102/batch", O_RDONLY| >> O_LARGEFILE) = 10 >> kfcntl(10, F_GETFD, 0x) = 0 >> kfcntl(10, F_SETFD, 0x0001)

Re: [sqlite] IOERR_DIR_FSYNC on main db

2008-07-14 Thread Dan
On Jul 15, 2008, at 4:46 AM, Ken wrote: > Preprocessor Macro: >#if defined (_AIX) > > Looks like the MJ file is not getting created, sqlite 3.5.9 This is > also reproducible via the command line: Possibly an issue with the > journal mode code that was introduced. > access("/hom

Re: [sqlite] IOERR_DIR_FSYNC on main db

2008-07-14 Thread Roger Binns
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Ken wrote: > open("/home/ixion/ix_propagator/data/db/ajax102/batch", O_RDONLY|O_LARGEFILE) > = 10 > kfcntl(10, F_GETFD, 0x) = 0 > kfcntl(10, F_SETFD, 0x0001) = 0 > fsync(10)

Re: [sqlite] IOERR_DIR_FSYNC on main db

2008-07-14 Thread Ken
Preprocessor Macro: #if defined (_AIX) Looks like the MJ file is not getting created, sqlite 3.5.9 This is also reproducible via the command line: Possibly an issue with the journal mode code that was introduced. SQLite version 3.5.9 Enter ".help" for instructions sqlite> attach da

Re: [sqlite] IOERR_DIR_FSYNC on main db

2008-07-14 Thread Ken
On the second commit I found that the locking_mode=exclusive caused the journal to merly be truncated vs deleted. I put the locking mode back to normal. It looks like the IOERR_DIR_FSYNC is due to the fact that a MJ journal is not being created! Pragma synchronous=OFF was in effect. The same

Re: [sqlite] IOERR_DIR_FSYNC on main db

2008-07-14 Thread Ken
Ok this is really weird. One sqlite Db functions fine. A second db (connection to a different db) commits and does not delete the journal file. I have verified that data is being written and the journal does indeed have the original blocks. Where in the source does sqlite make the decision to

Re: [sqlite] IOERR_DIR_FSYNC on main db

2008-07-14 Thread Ken
This might be an error on my side since the normal .db handling seems to function ok. The main .db is opened using sync_mode=off. So I shouldn't even get this particular error. I'll dig deeper into this and post more info once I validate the source of the error. John Stanton <[EMAIL PROTEC

Re: [sqlite] IOERR_DIR_FSYNC on main db

2008-07-14 Thread John Stanton
On earlier versions of AIX Sqlite has no issues. This problem might not be inherent to AIX V5.3. D. Richard Hipp wrote: > On Jul 14, 2008, at 1:31 PM, Ken wrote: > >> sqlite 3.5.9 and AIX 5.3 >> >> Upon commit of a main DB with an attached DB sqlite returns (1290) >> 0x50A SQLITE_IOERR_DIR

Re: [sqlite] IOERR_DIR_FSYNC on main db

2008-07-14 Thread D. Richard Hipp
On Jul 14, 2008, at 1:31 PM, Ken wrote: > sqlite 3.5.9 and AIX 5.3 > > Upon commit of a main DB with an attached DB sqlite returns (1290) > 0x50A SQLITE_IOERR_DIR_FSYNC > > > what would cause this? Any ideas on how to resolve? SQLite is attempting to fsync() a directory after unlinking the

[sqlite] IOERR_DIR_FSYNC on main db

2008-07-14 Thread Ken
sqlite 3.5.9 and AIX 5.3 Upon commit of a main DB with an attached DB sqlite returns (1290) 0x50A SQLITE_IOERR_DIR_FSYNC what would cause this? Any ideas on how to resolve? Thanks, Ken ___ sqlite-users mailing list sqlite-users@sqlite.org http