public inbox for [email protected]  
help / color / mirror / Atom feed
Hi Corruption in Postgres
2+ messages / 2 participants
[nested] [flat]

* Hi Corruption in Postgres
@ 2024-11-07 18:06  Pär Mattsson <[email protected]>
  0 siblings, 1 reply; 2+ messages in thread

From: Pär Mattsson @ 2024-11-07 18:06 UTC (permalink / raw)
  To: [email protected] <[email protected]>

Har någon typ av korruptionsproblematik sen 1/10 på lfv produktion :

pg_dump: error: Dumping the contents of table "aim_notam_originalpart" failed: PQgetResult() failed.
pg_dump: error: Error message from server: ERROR:  unexpected chunk size 1408 (expected 1661) in chunk 0 of 1 for toast value 18918695 in pg_toast_18194
pg_dump: error: The command was: COPY aim.aim_notam_originalpart (notam_id, part_num, created_at, e_details, canonical_filing_dtg, canonical_orig_addr, canonical_text) TO stdout;
aimdb_dump: pg_dump failed


1500000 rows



Mvh Pär


^ permalink  raw  reply  [nested|flat] 2+ messages in thread

* Re: Hi Corruption in Postgres
@ 2024-11-08 15:04  Achilleas Mantzios - cloud <[email protected]>
  parent: Pär Mattsson <[email protected]>
  0 siblings, 0 replies; 2+ messages in thread

From: Achilleas Mantzios - cloud @ 2024-11-08 15:04 UTC (permalink / raw)
  To: [email protected]


On 11/7/24 19:06, Pär Mattsson wrote:
>
> Har någon typ av korruptionsproblematik sen 1/10 på lfv produktion :
>
> pg_dump: error: Dumping the contents of table "aim_notam_originalpart" 
> failed: PQgetResult() failed.
> pg_dump: error: Error message from server: ERROR: unexpected chunk 
> size 1408 (expected 1661) in chunk 0 of 1 for toast value 18918695 in 
> pg_toast_18194
> pg_dump: error: The command was: COPY aim.aim_notam_originalpart 
> (notam_id, part_num, created_at, e_details, canonical_filing_dtg, 
> canonical_orig_addr, canonical_text) TO stdout;
> aimdb_dump: pg_dump failed
>
>
Your DB is corrupted. Either you have valid backup , or you try to 
rectify as is.

In this case, search the net / write a pgplsql func to pinpoint the 
corrupted toast row. Then zero it out.

This could be handy :

CREATE OR REPLACE FUNCTION find_bad_row(text,text,text) RETURNS VOID AS $f$

declare

tblname ALIAS FOR $1;

pkname ALIAS FOR $2;

culprit ALIAS FOR $3;

curid INT := 0;

vcontent TEXT;

badid INT;

begin

FOR badid IN EXECUTE 'SELECT '||pkname||' FROM '||tblname||' ORDER BY 
'||pkname LOOP

curid = curid + 1;

--raise notice 'inspecting row %', curid;

if curid % 100000 = 0 then

raise notice '% rows inspected', curid;

end if;

begin

EXECUTE 'SELECT '||culprit||' FROM '||tblname||' where '||pkname||' = 
'|| badid INTO vcontent;

vcontent := substr(vcontent,1,1000);

exception

when others then

raise notice 'data for table %,  pk=% is corrupt', tblname,badid;

continue;

end;

end loop;

end;

$f$ LANGUAGE plpgsql;

> 1500000 rows
>
>
>
> Mvh Pär

^ permalink  raw  reply  [nested|flat] 2+ messages in thread


end of thread, other threads:[~2024-11-08 15:04 UTC | newest]

Thread overview: 2+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2024-11-07 18:06 Hi Corruption in Postgres Pär Mattsson <[email protected]>
2024-11-08 15:04 ` Achilleas Mantzios - cloud <[email protected]>

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox