Re: [Info-Ingres] W_DM5061_WRONG_BLOB_LENGTH in verifydb



Oh and BTW. It is of course possible that there is a fault in the live
database and the recovery has simply reproduced this, but I can't do the
verify on the live database at the moment as it ties the table up for
way too long. I've managed to schedulle in the verify on live for this
Sunday....



Could be interesting.



Marty



From: info-ingres-bounces@xxxxxxxxxxxxxxxxxxxxxxxxx
[mailto:info-ingres-bounces@xxxxxxxxxxxxxxxxxxxxxxxxx] On Behalf Of
Martin Bowes
Sent: 18 September 2008 11:47
To: Ingres and related product discussion forum
Subject: [Info-Ingres] W_DM5061_WRONG_BLOB_LENGTH in verifydb



Hi All,



I'm running II 9.0.4 (a64.lnx/105)NPTL + patch12707.



I'm in the process of trialling the recovery of an 85G database to a
remote host. The recovery completed OK ...that is without signalling any
errors...but after the recovery I ran verifydb on all the tables in the
database. It has found one table which contains 4 blob columns (amongst
others) on which it is seriously unhappy.



Errors are of type:

W_DM5060_NO_ETAB_SEGMENT

W_DM505F_TOO_MANY_BAD_SEGS

W_DM5061_WRONG_BLOB_LENGTH WARNING: The coupon for attribute xb_innr
in record with tuple ID 4682756 contains a length of 153632, the
combined length of all it's segments is 61318.



I get 5 sets of these for 5 different tids.

column xb_innr: tids are 4682756, 43351554, 70878722

column xub_isr:tids are 40899589, 49575938



Note that the long columns in the table are:

xb_outr long varch not null not
default

xb_innr long varch not null not
default

xub_osr long varch with null with
default null

xub_isr long varch with null with
default null



Now the really fun thing....

When I check the table on the recovered database ...which has had no
work done on it...some of these tids are missing.

ie 4682756, 40899589 no longer exists in the table.



How....?



Here's another funny thing...

The verifydb ..-oxtables ultimately fails with the front end message:

Aborting because of error

E_QE0083 Error modifying a table.



In the errlog we see:

E_DM93AF_BAD_PAGE_CNT Wrong number of pages read or written. At page
112896, an attempt was made to read or write 16 pages, but only 0 pages
were actually processed.

write() failed with operating system error 0 (Success)

E_DM9006_BAD_FILE_WRITE Disk file write error on database:ace_pack_live
table:iietab_f3_abb
pathname:/dbdata4/II/ingres/data/default/ace_pack_live
filename:rabsi4ub.m03 page:112896

write() failed with operating system error 0 (Success)

E_DM9335_DM2F_ENDFILE End of file was reached during a read or write
operation.

E_DM92F0_DM1X_WRITE_GROUP Error writing a group of pages while
building a table or index.

E_DM92E5_DM1X_ALLOCATE Error allocating a page during build.

E_DM92E3_DM1X_NEWPAGE Error allocating new page for build.

E_DM925B_DM1B_BPUT Error occurred adding a record to a Btree table
being built.

E_DM9100_LOAD_TABLE_ERROR Error occurred while loading a table for a
modify, index, or load table operation.

E_DM0091_ERROR_MODIFYING_TABLE Error modifying a table.

E_QE0083_ERROR_MODIFYING_TABLE Error modifying a table.



Note thet the table iietab_f3_abb is associated with column xb_innr.



My monitoring of the data disk shows that it went 100% at this time. So
it looks like it was doing a modify. Does verifydb work by doing a
modify? If so the tids it reported would be completely scrambled by the
verifydb and would be useless to me when I try to track down the errant
data.



Anyone seen anything like this before?



BTW. There was a similar problem quite some time ago on II2.6 regarding
nullable blobs holding empty strings, but this is not the case here,
plus I reran the test case for that problem on my site and it came up
clean as a whistle.



Martin Bowes



Relevant Pages

  • [Info-Ingres] W_DM5061_WRONG_BLOB_LENGTH in verifydb
    ... I'm in the process of trialling the recovery of an 85G database to a ... The recovery completed OK ...that is without signalling any ... writefailed with operating system error 0 ... modify, index, or load table operation. ...
    (comp.databases.ingres)
  • Re: NTDS ISAM DataBase Corruption
    ... Did you tried the recover the corrupted Active Directory process that a gave ... This is brand new IBM Server. ... using symantec livestate recovery. ... Active Directory database is unavailable because it is damaged, ...
    (microsoft.public.windows.server.active_directory)
  • wow..
    ... The guy asks the way to DELETE the RECOVERY DATABASE FILE without dismounting anything or stopping any service. ... Is there anyway to either delete or mount this database without shutting ... down the store service and taking down my production databases? ...
    (microsoft.public.exchange.admin)
  • Re: Crash recovery after shutdown immediate?
    ... database stated it wanted a recovery. ... it seems that a crash recovery occured. ... ARCH: Archiving is disabled ... After that, I did a backup, NOT a restore. ...
    (comp.databases.oracle.server)
  • RE: Eseutil and restored database - Jet database errors
    ... database repair (hard recovery). ... It approximately spends 1 hour for 6~8 GB database. ... Please refer to the "Method 3: Restore a Mailbox from Backup to a Recovery ...
    (microsoft.public.exchange.admin)