Commit 54eff531 authored by Alvaro Herrera's avatar Alvaro Herrera

Fix deadlock hazard in CREATE INDEX CONCURRENTLY

Multiple sessions doing CREATE INDEX CONCURRENTLY simultaneously are
supposed to be able to work in parallel, as evidenced by fixes in commit
c3d09b3b specifically to support this case.  In reality, one of the
sessions would be aborted by a misterious "deadlock detected" error.

Jeff Janes diagnosed that this is because of leftover snapshots used for
system catalog scans -- this was broken by 8aa3e47510b9 keeping track of
(registering) the catalog snapshot.  To fix the deadlocks, it's enough
to de-register that snapshot prior to waiting.

Backpatch to 9.4, which introduced MVCC catalog scans.

Include an isolationtester spec that 8 out of 10 times reproduces the
deadlock with the unpatched code for me (Álvaro).

Author: Jeff Janes
Diagnosed-by: Jeff Janes
Reported-by: Jeremy Finzel
Discussion: https://postgr.es/m/CAMa1XUhHjCv8Qkx0WOr1Mpm_R4qxN26EibwCrj0Oor2YBUFUTg%40mail.gmail.com
parent 43803626
...@@ -856,11 +856,14 @@ DefineIndex(Oid relationId, ...@@ -856,11 +856,14 @@ DefineIndex(Oid relationId,
* doing CREATE INDEX CONCURRENTLY, which would see our snapshot as one * doing CREATE INDEX CONCURRENTLY, which would see our snapshot as one
* they must wait for. But first, save the snapshot's xmin to use as * they must wait for. But first, save the snapshot's xmin to use as
* limitXmin for GetCurrentVirtualXIDs(). * limitXmin for GetCurrentVirtualXIDs().
*
* Our catalog snapshot could have the same effect, so drop that one too.
*/ */
limitXmin = snapshot->xmin; limitXmin = snapshot->xmin;
PopActiveSnapshot(); PopActiveSnapshot();
UnregisterSnapshot(snapshot); UnregisterSnapshot(snapshot);
InvalidateCatalogSnapshot();
/* /*
* The index is now valid in the sense that it contains all currently * The index is now valid in the sense that it contains all currently
......
Parsed test spec with 2 sessions
starting permutation: s2l s1i s2i
step s2l: SELECT pg_advisory_lock(281457);
pg_advisory_lock
step s1i:
CREATE INDEX CONCURRENTLY mcic_one_pkey ON mcic_one (id)
WHERE lck_shr(281457);
<waiting ...>
step s2i:
CREATE INDEX CONCURRENTLY mcic_two_pkey ON mcic_two (id)
WHERE unlck();
step s1i: <... completed>
s1
...@@ -55,6 +55,7 @@ test: skip-locked-2 ...@@ -55,6 +55,7 @@ test: skip-locked-2
test: skip-locked-3 test: skip-locked-3
test: skip-locked-4 test: skip-locked-4
test: drop-index-concurrently-1 test: drop-index-concurrently-1
test: multiple-cic
test: alter-table-1 test: alter-table-1
test: alter-table-2 test: alter-table-2
test: alter-table-3 test: alter-table-3
......
# Test multiple CREATE INDEX CONCURRENTLY working simultaneously
setup
{
CREATE TABLE mcic_one (
id int
);
CREATE TABLE mcic_two (
id int
);
CREATE FUNCTION lck_shr(bigint) RETURNS bool IMMUTABLE LANGUAGE plpgsql AS $$
BEGIN PERFORM pg_advisory_lock_shared($1); RETURN true; END;
$$;
CREATE FUNCTION unlck() RETURNS bool IMMUTABLE LANGUAGE plpgsql AS $$
BEGIN PERFORM pg_advisory_unlock_all(); RETURN true; END;
$$;
}
teardown
{
DROP TABLE mcic_one, mcic_two;
DROP FUNCTION lck_shr(bigint);
DROP FUNCTION unlck();
}
session "s1"
step "s1i" {
CREATE INDEX CONCURRENTLY mcic_one_pkey ON mcic_one (id)
WHERE lck_shr(281457);
}
teardown { SELECT pg_advisory_unlock_all() AS "s1"; }
session "s2"
step "s2l" { SELECT pg_advisory_lock(281457); }
step "s2i" {
CREATE INDEX CONCURRENTLY mcic_two_pkey ON mcic_two (id)
WHERE unlck();
}
permutation "s2l" "s1i" "s2i"
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment