Thread: Select on pkey returned more than one row
Dear Gurus, As always, please point me to the right resource if this issue is already addressed. Version: 7.4.6, Debian "Woody" linux. Source: max func args increased to 64, but I doubt it matters. In a pl/pgsql function, called from an AFTER trigger, I receive this message: ERROR: query "SELECT mennyiseg FROM muvelet_cikk WHERE muvelet= $1 AND cikk= $2 AND minoseg= $3 AND mozgasnem= $4" returned more than one row Whereas the fields in the WHERE clause are exactly the pkey fields for that table: "muvelet_cikk_pkey" primary key, btree (muvelet, cikk, minoseg, mozgasnem) I'm not sure but this may be because of a race condition (one backend frequently inserting rows while another backend firing a trigger that causes a mass UPDATE on the same tuples. Indeed, we tend to receive deadlocks.) Are pkey constraints immediate or deferred? May it be the cause of the problem? If so, what are the risks of changing the constraint to immediate, if it can be done at all? The table definition in question is as follows. Double-checked, and indeed, no rows violate the pkey (and I assume no rows violate it in the transaction that throws the error, either -- but indeed, rows are updated or inserted into this table from both abovementioned backends) I'm able to send more information if needed. TIA, -- G. [local]:tir=# \d muvelet_cikk Table "public.muvelet_cikk" Column | Type | Modifiers -----------+---------------+-------------------- muvelet | integer | not null cikk | integer | not nullmozgasnem | integer | not null mennyiseg | numeric(14,4) | me | integer | kcikk | integer | minoseg | integer | not null default 1 Indexes: "muvelet_cikk_pkey" primary key, btree (muvelet, cikk, minoseg, mozgasnem) "muvelet_cikk_cikk_muvelet" btree(cikk, muvelet) "muvelet_cikk_mozgasnem_cikk" btree (mozgasnem, cikk) "muvelet_cikk_mozgasnem_muvelet" btree (mozgasnem,muvelet) Check constraints: "mennyiseg_me" CHECK ((mennyiseg IS NULL) = (me IS NULL)) Foreign-key constraints: "$2" FOREIGN KEY (cikk) REFERENCES cikk(az) ON UPDATE CASCADE "$6" FOREIGN KEY (kcikk) REFERENCESkulso_cikk(az) "$5" FOREIGN KEY (minoseg) REFERENCES minoseg(az) "$1" FOREIGN KEY (muvelet) REFERENCES muvelet(az)ON UPDATE CASCADE ON DELETE CASCADE "$4" FOREIGN KEY (me) REFERENCES mennyisegi_egyseg(az) "$3" FOREIGN KEY (mozgasnem) REFERENCES mozgasnem(az) Triggers: muvelet_cikk_ad AFTER DELETE ON muvelet_cikk FOR EACH ROW EXECUTE PROCEDURE muvelet_cikk_ad() muvelet_cikk_aiud AFTER INSERT OR DELETE OR UPDATE ON muvelet_cikk FOR EACH ROW EXECUTE PROCEDURE muvelet_cikk_aiud() muvelet_cikk_biu BEFORE INSERT OR UPDATE ON muvelet_cikk FOR EACH ROW EXECUTE PROCEDURE muvelet_cikk_biu() muvelet_cikk_noty AFTER INSERT OR DELETE OR UPDATE ON muvelet_cikk FOR EACH ROW EXECUTE PROCEDURE muvelet_cikk_noty()
Dear Gurus, I'm not sure if this mail reached the list members (at least, it reached the archive). If there's some formal problem that prevents delivery, please point me to the appropriate form's description. Yours, -- G. On 2005.06.28. 15:05, Szűcs Gábor wrote: > Dear Gurus, > > As always, please point me to the right resource if this issue is > already addressed. > > Version: 7.4.6, Debian "Woody" linux. > Source: max func args increased to 64, but I doubt it matters. > > In a pl/pgsql function, called from an AFTER trigger, I receive this > message: > > ERROR: query "SELECT mennyiseg FROM muvelet_cikk WHERE muvelet= $1 AND > cikk= $2 AND minoseg= $3 AND mozgasnem= $4 " returned more than one row > > Whereas the fields in the WHERE clause are exactly the pkey fields for > that table: > > "muvelet_cikk_pkey" primary key, btree (muvelet, cikk, minoseg, > mozgasnem) > > I'm not sure but this may be because of a race condition (one backend > frequently inserting rows while another backend firing a trigger that > causes a mass UPDATE on the same tuples. Indeed, we tend to receive > deadlocks.) > > Are pkey constraints immediate or deferred? May it be the cause of the > problem? If so, what are the risks of changing the constraint to > immediate, if it can be done at all? > > The table definition in question is as follows. Double-checked, and > indeed, no rows violate the pkey (and I assume no rows violate it in the > transaction that throws the error, either -- but indeed, rows are > updated or inserted into this table from both abovementioned backends) > > I'm able to send more information if needed. > > TIA, > -- > G. > > [local]:tir=# \d muvelet_cikk > Table "public.muvelet_cikk" > Column | Type | Modifiers > -----------+---------------+-------------------- > muvelet | integer | not null > cikk | integer | not null > mozgasnem | integer | not null > mennyiseg | numeric(14,4) | > me | integer | > kcikk | integer | > minoseg | integer | not null default 1 > Indexes: > "muvelet_cikk_pkey" primary key, btree (muvelet, cikk, minoseg, > mozgasnem) > "muvelet_cikk_cikk_muvelet" btree (cikk, muvelet) > "muvelet_cikk_mozgasnem_cikk" btree (mozgasnem, cikk) > "muvelet_cikk_mozgasnem_muvelet" btree (mozgasnem, muvelet) > Check constraints: > "mennyiseg_me" CHECK ((mennyiseg IS NULL) = (me IS NULL)) > Foreign-key constraints: > "$2" FOREIGN KEY (cikk) REFERENCES cikk(az) ON UPDATE CASCADE > "$6" FOREIGN KEY (kcikk) REFERENCES kulso_cikk(az) > "$5" FOREIGN KEY (minoseg) REFERENCES minoseg(az) > "$1" FOREIGN KEY (muvelet) REFERENCES muvelet(az) ON UPDATE CASCADE > ON DELETE CASCADE > "$4" FOREIGN KEY (me) REFERENCES mennyisegi_egyseg(az) > "$3" FOREIGN KEY (mozgasnem) REFERENCES mozgasnem(az) > Triggers: > muvelet_cikk_ad AFTER DELETE ON muvelet_cikk FOR EACH ROW EXECUTE > PROCEDURE muvelet_cikk_ad() > muvelet_cikk_aiud AFTER INSERT OR DELETE OR UPDATE ON muvelet_cikk > FOR EACH ROW EXECUTE PROCEDURE muvelet_cikk_aiud() > muvelet_cikk_biu BEFORE INSERT OR UPDATE ON muvelet_cikk FOR EACH > ROW EXECUTE PROCEDURE muvelet_cikk_biu() > muvelet_cikk_noty AFTER INSERT OR DELETE OR UPDATE ON muvelet_cikk > FOR EACH ROW EXECUTE PROCEDURE muvelet_cikk_noty() >
Szűcs Gábor <surrano@gmail.com> writes:
>> Version: 7.4.6, Debian "Woody" linux.
>> Source: max func args increased to 64, but I doubt it matters.
Well, it might, if you were sloppy about making sure that the software
had been completely rebuilt (including plpgsql and any other add-on
libraries you might be using). You generally want to do a "make clean"
and full rebuild after modifying any configuration values.
regards, tom lane