VACUUM was willing to remove a committed-dead tuple immediately if it was
deleted by the same transaction that inserted it. The idea is that such a
tuple could never have been visible to any other transaction, so we don't
need to keep it around to satisfy MVCC snapshots. However, there was
already an exception for tuples that are part of an update chain, and this
exception created a problem: we might remove TOAST tuples (which are never
part of an update chain) while their parent tuple stayed around (if it was
part of an update chain). This didn't pose a problem for most things,
since the parent tuple is indeed dead: no snapshot will ever consider it
visible. But MVCC-safe CLUSTER had a problem, since it will try to copy
RECENTLY_DEAD tuples to the new table. It then has to copy their TOAST
data too, and would fail if VACUUM had already removed the toast tuples.
Easiest fix is to get rid of the special case for xmin == xmax. This may
delay reclaiming dead space for a little bit in some cases, but it's by far
the most reliable way to fix the issue.
Per bug #5998 from Mark Reid. Back-patch to 8.3, which is the oldest
version with MVCC-safe CLUSTER.
With this patch, portals, SQL functions, and SPI all agree that there
should be only a CommandCounterIncrement between the queries that are
generated from a single SQL command by rule expansion. Fetching a whole
new snapshot now happens only between original queries. This is equivalent
to the existing behavior of EXPLAIN ANALYZE, and it was judged to be the
best choice since it eliminates one source of concurrency hazards for
rules. The patch should also make things marginally faster by reducing the
number of snapshot push/pop operations.
The patch removes pg_parse_and_rewrite(), which is no longer used anywhere.
There was considerable discussion about more aggressive refactoring of the
query-processing functions exported by postgres.c, but for the moment
nothing more has been done there.
I also took the opportunity to refactor snapmgr.c's API slightly: the
former PushUpdatedSnapshot() has been split into two functions.
Marko Tiikkaja, reviewed by Steve Singer and Tom Lane
Until now, our Serializable mode has in fact been what's called Snapshot
Isolation, which allows some anomalies that could not occur in any
serialized ordering of the transactions. This patch fixes that using a
method called Serializable Snapshot Isolation, based on research papers by
Michael J. Cahill (see README-SSI for full references). In Serializable
Snapshot Isolation, transactions run like they do in Snapshot Isolation,
but a predicate lock manager observes the reads and writes performed and
aborts transactions if it detects that an anomaly might occur. This method
produces some false positives, ie. it sometimes aborts transactions even
though there is no anomaly.
To track reads we implement predicate locking, see storage/lmgr/predicate.c.
Whenever a tuple is read, a predicate lock is acquired on the tuple. Shared
memory is finite, so when a transaction takes many tuple-level locks on a
page, the locks are promoted to a single page-level lock, and further to a
single relation level lock if necessary. To lock key values with no matching
tuple, a sequential scan always takes a relation-level lock, and an index
scan acquires a page-level lock that covers the search key, whether or not
there are any matching keys at the moment.
A predicate lock doesn't conflict with any regular locks or with another
predicate locks in the normal sense. They're only used by the predicate lock
manager to detect the danger of anomalies. Only serializable transactions
participate in predicate locking, so there should be no extra overhead for
for other transactions.
Predicate locks can't be released at commit, but must be remembered until
all the transactions that overlapped with it have completed. That means that
we need to remember an unbounded amount of predicate locks, so we apply a
lossy but conservative method of tracking locks for committed transactions.
If we run short of shared memory, we overflow to a new "pg_serial" SLRU
pool.
We don't currently allow Serializable transactions in Hot Standby mode.
That would be hard, because even read-only transactions can cause anomalies
that wouldn't otherwise occur.
Serializable isolation mode now means the new fully serializable level.
Repeatable Read gives you the old Snapshot Isolation level that we have
always had.
Kevin Grittner and Dan Ports, reviewed by Jeff Davis, Heikki Linnakangas and
Anssi Kääriäinen
transaction snapshots, i.e. a snapshot registered at the beginning of
a transaction. Change variable naming and comments to reflect this reality
in preparation for a future, truly serializable mode, e.g.
Serializable Snapshot Isolation (SSI).
For the moment transaction snapshots are still used to implement
SERIALIZABLE, but hopefully not for too much longer. Patch by Kevin
Grittner and Dan Ports with review and some minor wording changes by me.
VACUUM FULL INPLACE), along with a boatload of subsidiary code and complexity.
Per discussion, the use case for this method of vacuuming is no longer large
enough to justify maintaining it; not to mention that we don't wish to invest
the work that would be needed to make it play nicely with Hot Standby.
Aside from the code directly related to old-style VACUUM FULL, this commit
removes support for certain WAL record types that could only be generated
within VACUUM FULL, redirect-pointer removal in heap_page_prune, and
nontransactional generation of cache invalidation sinval messages (the last
being the sticking point for Hot Standby).
We still have to retain all code that copes with finding HEAP_MOVED_OFF and
HEAP_MOVED_IN flag bits on existing tuples. This can't be removed as long
as we want to support in-place update from pre-9.0 databases.
Enabled by recovery_connections = on (default) and forcing archive recovery using a recovery.conf. Recovery processing now emulates the original transactions as they are replayed, providing full locking and MVCC behaviour for read only queries. Recovery must enter consistent state before connections are allowed, so there is a delay, typically short, before connections succeed. Replay of recovering transactions can conflict and in some cases deadlock with queries during recovery; these result in query cancellation after max_standby_delay seconds have expired. Infrastructure changes have minor effects on normal running, though introduce four new types of WAL record.
New test mode "make standbycheck" allows regression tests of static command behaviour on a standby server while in recovery. Typical and extreme dynamic behaviours have been checked via code inspection and manual testing. Few port specific behaviours have been utilised, though primary testing has been on Linux only so far.
This commit is the basic patch. Additional changes will follow in this release to enhance some aspects of behaviour, notably improved handling of conflicts, deadlock detection and query cancellation. Changes to VACUUM FULL are also required.
Simon Riggs, with significant and lengthy review by Heikki Linnakangas, including streamlined redesign of snapshot creation and two-phase commit.
Important contributions from Florian Pflug, Mark Kirkwood, Merlin Moncure, Greg Stark, Gianni Ciolli, Gabriele Bartolini, Hannu Krosing, Robert Haas, Tatsuo Ishii, Hiroyuki Yamada plus support and feedback from many other community members.
Partially revert the previous patch I installed and replace it with a more
general fix: any time a snapshot is pushed as Active, we need to ensure that it
will not be modified in the future. This means that if the same snapshot is
used as CurrentSnapshot, it needs to be copied separately. This affects
serializable transactions only, because CurrentSnapshot has already been copied
by RegisterSnapshot and so PushActiveSnapshot does not think it needs another
copy. However, CommandCounterIncrement would modify CurrentSnapshot, whereas
ActiveSnapshots must not have their command counters incremented.
I say "partially" because the regression test I added for the previous bug
has been kept.
(This restores 8.3 behavior, because before snapmgr.c existed, any snapshot set
as Active was copied.)
Per bug report from Stuart Bishop in
6bc73d4c0910042358k3d1adff3qa36f8df75198ecea@mail.gmail.com
The old coding was using a regular snapshot, referenced elsewhere, that was
subject to having its command counter updated. Fix by creating a private copy
of the snapshot exclusively for the cursor.
Backpatch to 8.4, which is when the bug was introduced during the snapshot
management rewrite.
non-writable large objects need to have their snapshots registered on the
transaction resowner, not the current portal's, because it must persist until
the large object is closed (which the portal does not). Also, ensure that the
serializable snapshot is recorded by the transaction resource owner too, even
when a subtransaction has changed the current resource owner before
serializable is taken.
Per bug reports from Pavan Deolasee.
by hand. As an added bonus, the new code is smaller and more understandable,
and the ugly loops are gone.
This had been discussed all along but never implemented. It became clear that
it really needed to be fixed after a bug report by Pavan Deolasee.
GetOldestXmin() instead of RecentGlobalXmin; this is safer because we do not
depend on the latter being correctly set elsewhere, and while it is more
expensive, this code path is not performance-critical. This is a real
risk for autovacuum, because it can execute whole cycles without doing
a single vacuum, which would mean that RecentGlobalXmin would stay at its
initialization value, FirstNormalTransactionId, causing a bogus value to be
inserted in pg_database. This bug could explain some recent reports of
failure to truncate pg_clog.
At the same time, change the initialization of RecentGlobalXmin to
InvalidTransactionId, and ensure that it's set to something else whenever
it's going to be used. Using it as FirstNormalTransactionId in HOT page
pruning could incur in data loss. InitPostgres takes care of setting it
to a valid value, but the extra checks are there to prevent "special"
backends from behaving in unusual ways.
Per Tom Lane's detailed problem dissection in 29544.1221061979@sss.pgh.pa.us
command id is the cmin, when it can in fact be a combo cid. That made rows
incorrectly invisible to a transaction where a tuple was deleted by multiple
aborted subtransactions.
Report and patch Karl Schnaitter. Back-patch to 8.3, where combo cids was
introduced.
CopySnapshot, per Neil Conway. Also add a comment about the assumption in
GetSnapshotData that the argument is statically allocated.
Also, fix some more typos in comments in snapmgr.c.
There are two ways to track a snapshot: there's the "registered" list, which
is used for arbitrary long-lived snapshots; and there's the "active stack",
which is used for the snapshot that is considered "active" at any time.
This also allows users of snapshots to stop worrying about snapshot memory
allocation and freeing, and about using PG_TRY blocks around ActiveSnapshot
assignment. This is all done automatically now.
As a consequence, this allows us to reset MyProc->xmin when there are no
more snapshots registered in the current backend, reducing the impact that
long-running transactions have on VACUUM.
snapmgmt.c file for the former. The header files have also been reorganized
in three parts: the most basic snapshot definitions are now in a new file
snapshot.h, and the also new snapmgmt.h keeps the definitions for snapmgmt.c.
tqual.h has been reduced to the bare minimum.
This patch is just a first step towards managing live snapshots within a
transaction; there is no functionality change.
Per my proposal to pgsql-patches on 20080318191940.GB27458@alvh.no-ip.org and
subsequent discussion.
but no database changes have been made since the last CommandCounterIncrement.
This should result in a significant improvement in the number of "commands"
that can typically be performed within a transaction before hitting the 2^32
CommandId size limit. In particular this buys back (and more) the possible
adverse consequences of my previous patch to fix plan caching behavior.
The implementation requires tracking whether the current CommandCounter
value has been "used" to mark any tuples. CommandCounter values stored into
snapshots are presumed not to be used for this purpose. This requires some
small executor changes, since the executor used to conflate the curcid of
the snapshot it was using with the command ID to mark output tuples with.
Separating these concepts allows some small simplifications in executor APIs.
Something for the TODO list: look into having CommandCounterIncrement not do
AcceptInvalidationMessages. It seems fairly bogus to be doing it there,
but exactly where to do it instead isn't clear, and I'm disinclined to mess
with asynchronous behavior during late beta.
TransactionIdDidAbort, when handling the case that xmin is one of the current
transaction's XIDs and the tuple has been deleted. xmax must also be one of
the current transaction's XIDs, since no one else can see it yet, and it's
cheaper to look at local state than shared state to find out if xmax aborted.
Per an idea of Heikki's.
ReadNewTransactionId from GetSnapshotData --- with a "latestCompletedXid"
variable that is updated during transaction commit or abort. Since
latestCompletedXid is written only in places that had to lock ProcArrayLock
exclusively anyway, and is read only in places that had to lock ProcArrayLock
shared anyway, it adds no new locking requirements to the system despite being
cluster-wide. Moreover, removing ReadNewTransactionId from snapshot
acquisition eliminates the need to take both XidGenLock and ProcArrayLock at
the same time. Since XidGenLock is sometimes held across I/O this can be a
significant win. Some preliminary benchmarking suggested that this patch has
no effect on average throughput but can significantly improve the worst-case
transaction times seen in pgbench. Concept by Florian Pflug, implementation
by Tom Lane.
that still thought they could set HEAP_XMAX_COMMITTED immediately after
seeing the other transaction commit. Make them use the same logic as
tqual.c does to determine if the hint bit can be set yet.
before reporting a transaction committed. Data consistency is still
guaranteed (unlike setting fsync = off), but a crash may lose the effects
of the last few transactions. Patch by Simon, some editorialization by Tom.
pointer" in every Snapshot struct. This allows removal of the case-by-case
tests in HeapTupleSatisfiesVisibility, which should make it a bit faster
(I didn't try any performance tests though). More importantly, we are no
longer violating portable C practices by assuming that small integers are
distinct from all pointer values, and HeapTupleSatisfiesDirty no longer
has a non-reentrant API involving side-effects on a global variable.
There were a couple of places calling HeapTupleSatisfiesXXX routines
directly rather than through the HeapTupleSatisfiesVisibility macro.
Since these places had to be changed anyway, I chose to make them go
through the macro for uniformity.
Along the way I renamed HeapTupleSatisfiesSnapshot to HeapTupleSatisfiesMVCC
to emphasize that it's only used with MVCC-type snapshots. I was sorely
tempted to rename HeapTupleSatisfiesVisibility to HeapTupleSatisfiesSnapshot,
but forebore for the moment to avoid confusion and reduce the likelihood that
this patch breaks some of the pending patches. Might want to reconsider
doing that later.
keeping private state in each backend that has inserted and deleted the same
tuple during its current top-level transaction. This is sufficient since
there is no need to be able to determine the cmin/cmax from any other
transaction. This gets us back down to 23-byte headers, removing a penalty
paid in 8.0 to support subtransactions. Patch by Heikki Linnakangas, with
minor revisions by moi, following a design hashed out awhile back on the
pghackers list.
cases where we already hold the desired lock "indirectly", either via
membership in a MultiXact or because the lock was originally taken by a
different subtransaction of the current transaction. These cases must be
accounted for to avoid needless deadlocks and/or inappropriate replacement of
an exclusive lock with a shared lock. Per report from Clarence Gardner and
subsequent investigation.
in PITR scenarios. We now WAL-log the replacement of old XIDs with
FrozenTransactionId, so that such replacement is guaranteed to propagate to
PITR slave databases. Also, rather than relying on hint-bit updates to be
preserved, pg_clog is not truncated until all instances of an XID are known to
have been replaced by FrozenTransactionId. Add new GUC variables and
pg_autovacuum columns to allow management of the freezing policy, so that
users can trade off the size of pg_clog against the amount of freezing work
done. Revise the already-existing code that forces autovacuum of tables
approaching the wraparound point to make it more bulletproof; also, revise the
autovacuum logic so that anti-wraparound vacuuming is done per-table rather
than per-database. initdb forced because of changes in pg_class, pg_database,
and pg_autovacuum catalogs. Heikki Linnakangas, Simon Riggs, and Tom Lane.
PGPROC array into snapshots, and use this information to avoid visits
to pg_subtrans in HeapTupleSatisfiesSnapshot. This appears to solve
the pg_subtrans-related context swap storm problem that's been reported
by several people for 8.1. While at it, modify GetSnapshotData to not
take an exclusive lock on ProcArrayLock, as closer analysis shows that
shared lock is always sufficient.
Itagaki Takahiro and Tom Lane
comment line where output as too long, and update typedefs for /lib
directory. Also fix case where identifiers were used as variable names
in the backend, but as typedefs in ecpg (favor the backend for
indenting).
Backpatch to 8.1.X.
