My company's product is developing a product using postgres 7.4.6 on a
Linux 2.6.10 or 11 kernel, (don't have the exact version handy). We
do "vacuum verbose analyze" daily. Two days after the database was
created and loaded with a small amount of data (megabytes), the
postgres logs had this line:
2005-04-16 00:00:53 ERROR: could not read block 52 of
relation "idx_ef_hash": Input/output error
idx_ef_hash is an index. I believe that vacuuming is to blame based on
vacuum output:
04/16 00:00:14.388 03136277 Starting daily vacuum
04/16 00:00:53.192 03136280 process failed with exit code 1.
04/16 00:00:53.193 03136281 stderr: INFO: vacuuming
"information_schema.sql_features"
04/16 00:00:53.193 03136282 stderr: INFO: "sql_features": found 0
removable, 360 nonremovable row versions in 5 pages
04/16 00:00:53.193 03136283 stderr: DETAIL: 0 dead row versions cannot be
removed yet.
04/16 00:00:53.193 03136284 stderr: There were 0 unused item pointers.
04/16 00:00:53.193 03136285 stderr: 0 pages are entirely empty.
04/16 00:00:53.193 03136286 stderr: CPU 0.00s/0.00u sec elapsed 0.00 sec.
04/16 00:00:53.193 03136287 stderr: INFO: vacuuming
"pg_toast.pg_toast_17070"
...
(The first two lines are from my application, the rest is stderr from
the vacuum process.)
Googling didn't turn up useful information on this problem. Can anyone
tell me whether this error indicates a disk problem, OS problem, known
postgres bug, or something else? We've only seen this occur once.
Jack Orenstein
----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.