Re: missing chunk 0 for toast value ...

On Tue, Oct 25, 2011 at 3:47 PM, Andrew Hammond
<andrew(dot)george(dot)hammond(at)gmail(dot)com> wrote:
> I found the following error message in my logfiles.
> Oct 24 04:05:57 db-app02 postgres[24640]: [2-1]
> user=nutricate,db=admin_production ERROR:  missing chunk number 0
> for toast value 2411466 in pg_toast_2619
> Oct 24 04:05:57 db-app02 postgres[24640]: [2-2]
> user=nutricate,db=admin_production STATEMENT:  SELECT "devices"."id",
> "devices"."name", "devices"."info", "devices"."pos_id",
> "devices"."device_group_id", "devices"."header_id", "devices"."footer_id",
> "devices"."device_type_id", "devices"."auth_code", "devices"."pos_comm_id",
> "devices"."printer_comm_id", "devices"."sw_version", "devices"."hw_version",
> "devices"."device_status", "devices"."entity_id", "devices"."address",
> "devices"."created_by", "devices"."create_method", "devices"."created_on",
> "devices"."modified_by", "devices"."updated_on", "devices"."version_id",
> "devices"."unique_id", "devices"."hw_key", "devices"."config_status",
> "devices"."activated_on", "devices"."store_id",
> "devices"."last_status_update", "devices"."loaded_on",
> "devices"."header2_id", "devices"."footer2_id", "devices"."timezone_id",
> "devices"."scanner_comm_id", "devices"."public_address",
> "devices"."hostname", "devices"."update_sw_version",
> "devices"."proxy_address", "devices"."proxy_type_id",
> "devices"."build_error",
> Oct 24 04:05:57 db-app02 postgres[24640]: [2-3]  "devices"."local_info",
> "devices"."associated_on" FROM "devices" WHERE ("devices"."entity_id" IN
> (SELECT U0."id" FROM "entities" U0 WHERE ((U0."lft" > 280  AND U0."rgt" <
> 2597 ) OR U0."id" = 140 )) AND "devices"."auth_code" = E'0063' )
> I tried reproducing it with
> SELECT * FROM devices WHERE ("devices"."entity_id" IN (SELECT U0."id" FROM
> "entities" U0 WHERE ((U0."lft" > 280  AND U0."rgt" < 2597 ) OR U0."id" = 140
> )) AND "devices"."auth_code" = E'0063' )
> This returned no rows and did not reproduce the error. I tried reproducing
> with an even broader approach
> CREATE TEMP TABLE foo AS SELECT * FROM devices;
> CREATE TEMP TABLE bar AS SELECT * FROM entities;
> Still no repro. This is the second time I've seen this error, and the last
> time it also did not reproduce. I'm wondering a couple of things. First, is
> there some other way to reproduce this error? Second, would simply deleting
> the row cause the problem to go away? I wouldn't think so, but why else is
> it not reproducing? Third, what is the recommended course of action here. We
> have hourly backups, but this is a production database, so I would prefer
> not to have to restore. I tested a restore from an hour ago and it ran
> successfully.

hrm, that's bizarre -- version etc?

merlin