When marking a branch as half-dead, a pointer to the top of the branch is
stored in the leaf block's hi-key. During normal operation, the high key
was left in place, and the block number was just stored in the ctid field
of the high key tuple, but in WAL replay, the high key was recreated as a
truncated tuple with zero columns. For the sake of easier debugging, also
truncate the tuple in normal operation, so that the page is identical
after WAL replay. Also, rename the 'downlink' field in the WAL record to
'topparent', as that seems like a more descriptive name. And make sure
it's set to invalid when unlinking the leaf page.
Forgot to update LSN of left sibling's page, when creating a new root.
I fixed this for regular insertions and page splits earlier, but missed
new root creation.
Inserting a downlink to an internal page clears the incomplete-split flag
of the child's left sibling, so the left sibling's LSN also needs to be
updated.
With this in place, a session blocking behind another one because of
tuple locks will get a context line mentioning the relation name, tuple
TID, and operation being done on tuple. For example:
LOG: process 11367 still waiting for ShareLock on transaction 717 after 1000.108 ms
DETAIL: Process holding the lock: 11366. Wait queue: 11367.
CONTEXT: while updating tuple (0,2) in relation "foo"
STATEMENT: UPDATE foo SET value = 3;
Most usefully, the new line is displayed by log entries due to
log_lock_waits, although of course it will be printed by any other log
message as well.
Author: Christian Kruse, some tweaks by Álvaro Herrera
Reviewed-by: Amit Kapila, Andres Freund, Tom Lane, Robert Haas
Splitting a page consists of two separate steps: splitting the child page,
and inserting the downlink for the new right page to the parent. Previously,
we handled the case that you crash in between those steps with a cleanup
routine after the WAL recovery had finished, which finished the incomplete
split. However, that doesn't help if the page split is interrupted but the
database doesn't crash, so that you don't perform WAL recovery. That could
happen for example if you run out of disk space.
Remove the end-of-recovery cleanup step. Instead, when a page is split, the
left page is marked with a new INCOMPLETE_SPLIT flag, and when the downlink
is inserted to the parent, the flag is cleared again. If an insertion sees
a page with the flag set, it knows that the split was interrupted for some
reason, and inserts the missing downlink before proceeding.
I used the same approach to fix GIN and GiST split algorithms earlier. This
was the last WAL cleanup routine, so we could get rid of that whole
machinery now, but I'll leave that for a separate patch.
Reviewed by Peter Geoghegan.
In short, we don't allow a page to be deleted if it's the rightmost child
of its parent, but that situation can change after we check for it.
Problem
-------
We check that the page to be deleted is not the rightmost child of its
parent, and then lock its left sibling, the page itself, its right sibling,
and the parent, in that order. However, if the parent page is split after
the check but before acquiring the locks, the target page might become the
rightmost child, if the split happens at the right place. That leads to an
error in vacuum (I reproduced this by setting a breakpoint in debugger):
ERROR: failed to delete rightmost child 41 of block 3 in index "foo_pkey"
We currently re-check that the page is still the rightmost child, and throw
the above error if it's not. We could easily just give up rather than throw
an error, but that approach doesn't scale to half-dead pages. To recap,
although we don't normally allow deleting the rightmost child, if the page
is the *only* child of its parent, we delete the child page and mark the
parent page as half-dead in one atomic operation. But before we do that, we
check that the parent can later be deleted, by checking that it in turn is
not the rightmost child of the grandparent (potentially recursing all the
way up to the root). But the same situation can arise there - the
grandparent can be split while we're not holding the locks. We end up with
a half-dead page that we cannot delete.
To make things worse, the keyspace of the deleted page has already been
transferred to its right sibling. As the README points out, the keyspace at
the grandparent level is "out-of-whack" until the half-dead page is deleted,
and if enough tuples with keys in the transferred keyspace are inserted, the
page might get split and a downlink might be inserted into the grandparent
that is out-of-order. That might not cause any serious problem if it's
transient (as the README ponders), but is surely bad if it stays that way.
Solution
--------
This patch changes the page deletion algorithm to avoid that problem. After
checking that the topmost page in the chain of to-be-deleted pages is not
the rightmost child of its parent, and then deleting the pages from bottom
up, unlink the pages from top to bottom. This way, the intermediate stages
are similar to the intermediate stages in page splitting, and there is no
transient stage where the keyspace is "out-of-whack". The topmost page in
the to-be-deleted chain doesn't have a downlink pointing to it, like a page
split before the downlink has been inserted.
This also allows us to get rid of the cleanup step after WAL recovery, if we
crash during page deletion. The deletion will be continued at next VACUUM,
but the tree is consistent for searches and insertions at every step.
This bug is old, all supported versions are affected, but this patch is too
big to back-patch (and changes the WAL record formats of related records).
We have not heard any reports of the bug from users, so clearly it's not
easy to bump into. Maybe backpatch later, after this has had some field
testing.
Reviewed by Kevin Grittner and Peter Geoghegan.
These flushes were added in my commit d2896a9ed, which added the btree
logic that keeps a cached copy of the index metapage data in index relcache
entries. The idea was to ensure that other backends would promptly update
their cached copies after a change. However, this is not really necessary,
since _bt_getroot() has adequate defenses against believing a stale root
page link, and _bt_getrootheight() doesn't have to be 100% right.
Moreover, if it were necessary, a relcache flush would be an unreliable way
to do it, since the sinval mechanism believes that relcache flush requests
represent transactional updates, and therefore discards them on transaction
rollback. Therefore, we might as well drop these flush requests and save
the time to rebuild the whole relcache entry after a metapage change.
If we ever try to support in-place truncation of btree indexes, it might
be necessary to revisit this issue so that _bt_getroot() can't get caught
by trying to follow a metapage link to a page that no longer exists.
A possible solution to that is to make use of an smgr, rather than
relcache, inval request to force other backends to discard their cached
metapages. But for the moment this is not worth pursuing.
Since C99, it's been standard for printf and friends to accept a "z" size
modifier, meaning "whatever size size_t has". Up to now we've generally
dealt with printing size_t values by explicitly casting them to unsigned
long and using the "l" modifier; but this is really the wrong thing on
platforms where pointers are wider than longs (such as Win64). So let's
start using "z" instead. To ensure we can do that on all platforms, teach
src/port/snprintf.c to understand "z", and add a configure test to force
use of that implementation when the platform's version doesn't handle "z".
Having done that, modify a bunch of places that were using the
unsigned-long hack to use "z" instead. This patch doesn't pretend to have
gotten everyplace that could benefit, but it catches many of them. I made
an effort in particular to ensure that all uses of the same error message
text were updated together, so as not to increase the number of
translatable strings.
It's possible that this change will result in format-string warnings from
pre-C99 compilers. We might have to reconsider if there are any popular
compilers that will warn about this; but let's start by seeing what the
buildfarm thinks.
Andres Freund, with a little additional work by me
In ordinary operation, VACUUM must be careful to take a cleanup lock on
each leaf page of a btree index; this ensures that no indexscans could
still be "in flight" to heap tuples due to be deleted. (Because of
possible index-tuple motion due to concurrent page splits, it's not enough
to lock only the pages we're deleting index tuples from.) In Hot Standby,
the WAL replay process must likewise lock every leaf page. There were
several bugs in the code for that:
* The replay scan might come across unused, all-zero pages in the index.
While btree_xlog_vacuum itself did the right thing (ie, nothing) with
such pages, xlogutils.c supposed that such pages must be corrupt and
would throw an error. This accounts for various reports of replication
failures with "PANIC: WAL contains references to invalid pages". To
fix, add a ReadBufferMode value that instructs XLogReadBufferExtended
not to complain when we're doing this.
* btree_xlog_vacuum performed the extra locking if standbyState ==
STANDBY_SNAPSHOT_READY, but that's not the correct test: we won't open up
for hot standby queries until the database has reached consistency, and
we don't want to do the extra locking till then either, for fear of reading
corrupted pages (which bufmgr.c would complain about). Fix by exporting a
new function from xlog.c that will report whether we're actually in hot
standby replay mode.
* To ensure full coverage of the index in the replay scan, btvacuumscan
would emit a dummy WAL record for the last page of the index, if no
vacuuming work had been done on that page. However, if the last page
of the index is all-zero, that would result in corruption of said page,
since the functions called on it weren't prepared to handle that case.
There's no need to lock any such pages, so change the logic to target
the last normal leaf page instead.
The first two of these bugs were diagnosed by Andres Freund, the other one
by me. Fixes based on ideas from Heikki Linnakangas and myself.
This has been wrong since Hot Standby was introduced, so back-patch to 9.0.
SnapshotNow scans have the undesirable property that, in the face of
concurrent updates, the scan can fail to see either the old or the new
versions of the row. In many cases, we work around this by requiring
DDL operations to hold AccessExclusiveLock on the object being
modified; in some cases, the existing locking is inadequate and random
failures occur as a result. This commit doesn't change anything
related to locking, but will hopefully pave the way to allowing lock
strength reductions in the future.
The major issue has held us back from making this change in the past
is that taking an MVCC snapshot is significantly more expensive than
using a static special snapshot such as SnapshotNow. However, testing
of various worst-case scenarios reveals that this problem is not
severe except under fairly extreme workloads. To mitigate those
problems, we avoid retaking the MVCC snapshot for each new scan;
instead, we take a new snapshot only when invalidation messages have
been processed. The catcache machinery already requires that
invalidation messages be sent before releasing the related heavyweight
lock; else other backends might rely on locally-cached data rather
than scanning the catalog at all. Thus, making snapshot reuse
dependent on the same guarantees shouldn't break anything that wasn't
already subtly broken.
Patch by me. Review by Michael Paquier and Andres Freund.
MarkBufferDirtyHint() writes WAL, and should know if it's got a
standard buffer or not. Currently, the only callers where buffer_std
is false are related to the FSM.
In passing, rename XLOG_HINT to XLOG_FPI, which is more descriptive.
Back-patch to 9.3.
Checksums are set immediately prior to flush out of shared buffers
and checked when pages are read in again. Hint bit setting will
require full page write when block is dirtied, which causes various
infrastructure changes. Extensive comments, docs and README.
WARNING message thrown if checksum fails on non-all zeroes page;
ERROR thrown but can be disabled with ignore_checksum_failure = on.
Feature enabled by an initdb option, since transition from option off
to option on is long and complex and has not yet been implemented.
Default is not to use checksums.
Checksum used is WAL CRC-32 truncated to 16-bits.
Simon Riggs, Jeff Davis, Greg Smith
Wide input and assistance from many community members. Thank you.
Remove use of PageSetTLI() from all page manipulation functions
and adjust README to indicate change in the way we make changes
to pages. Repurpose those bytes into the pd_checksum field and
explain how that works in comments about page header.
Refactoring ahead of actual feature patch which would make use
of the checksum field, arriving later.
Jeff Davis, with comments and doc changes by Simon Riggs
Direction suggested by Robert Haas; many others providing
review comments.
This patch addresses the problem that applications currently have to
extract object names from possibly-localized textual error messages,
if they want to know for example which index caused a UNIQUE_VIOLATION
failure. It adds new error message fields to the wire protocol, which
can carry the name of a table, table column, data type, or constraint
associated with the error. (Since the protocol spec has always instructed
clients to ignore unrecognized field types, this should not create any
compatibility problem.)
Support for providing these new fields has been added to just a limited set
of error reports (mainly, those in the "integrity constraint violation"
SQLSTATE class), but we will doubtless add them to more calls in future.
Pavel Stehule, reviewed and extensively revised by Peter Geoghegan, with
additional hacking by Tom Lane.
Historically we've used a couple of very ad-hoc fudge factors to try to
get the right results when indexes of different sizes would satisfy a
query with the same number of index leaf tuples being visited. In
commit 21a39de580 I tweaked one of these
fudge factors, with results that proved disastrous for larger indexes.
Commit bf01e34b55 fudged it some more,
but still with not a lot of principle behind it.
What seems like a better way to address these issues is to explicitly model
index-descent costs, since that's what's really at stake when considering
diferent indexes with similar leaf-page-level costs. We tried that once
long ago, and found that charging random_page_cost per page descended
through was way too much, because upper btree levels tend to stay in cache
in real-world workloads. However, there's still CPU costs to think about,
and the previous fudge factors can be seen as a crude attempt to account
for those costs. So this patch replaces those fudge factors with explicit
charges for the number of tuple comparisons needed to descend the index
tree, plus a small charge per page touched in the descent. The cost
multipliers are chosen so that the resulting charges are in the vicinity of
the historical (pre-9.2) fudge factors for indexes of up to about a million
tuples, while not ballooning unreasonably beyond that, as the old fudge
factor did (even more so in 9.2).
To make this work accurately for btree indexes, add some code that allows
extraction of the known root-page height from a btree. There's no
equivalent number readily available for other index types, but we can use
the log of the number of index pages as an approximate substitute.
This seems like too much of a behavioral change to risk back-patching,
but it should improve matters going forward. In 9.2 I'll just revert
the fudge-factor change.
This gets rid of XLByteLT, XLByteLE, XLByteEQ and XLByteAdvance.
These were useful for brevity when XLogRecPtrs were split in
xlogid/xrecoff; but now that they are simple uint64's, they are just
clutter. The only downside to making this change would be ease of
backporting patches, but that has been negated by other substantive
changes to the involved code anyway. The clarity of simpler expressions
makes the change worthwhile.
Most of the changes are mechanical, but in a couple of places, the patch
author chose to invert the operator sense, making the code flow more
logical (and more in line with preceding comments).
Author: Andres Freund
Eyeballed by Dimitri Fontaine and Alvaro Herrera
Most of the replay functions for WAL record types that modify more than
one page failed to ensure that those pages were locked correctly to ensure
that concurrent queries could not see inconsistent page states. This is
a hangover from coding decisions made long before Hot Standby was added,
when it was hardly necessary to acquire buffer locks during WAL replay
at all, let alone hold them for carefully-chosen periods.
The key problem was that RestoreBkpBlocks was written to hold lock on each
page restored from a full-page image for only as long as it took to update
that page. This was guaranteed to break any WAL replay function in which
there was any update-ordering constraint between pages, because even if the
nominal order of the pages is the right one, any mixture of full-page and
non-full-page updates in the same record would result in out-of-order
updates. Moreover, it wouldn't work for situations where there's a
requirement to maintain lock on one page while updating another. Failure
to honor an update ordering constraint in this way is thought to be the
cause of bug #7648 from Daniel Farina: what seems to have happened there
is that a btree page being split was rewritten from a full-page image
before the new right sibling page was written, and because lock on the
original page was not maintained it was possible for hot standby queries to
try to traverse the page's right-link to the not-yet-existing sibling page.
To fix, get rid of RestoreBkpBlocks as such, and instead create a new
function RestoreBackupBlock that restores just one full-page image at a
time. This function can be invoked by WAL replay functions at the points
where they would otherwise perform non-full-page updates; in this way, the
physical order of page updates remains the same no matter which pages are
replaced by full-page images. We can then further adjust the logic in
individual replay functions if it is necessary to hold buffer locks
for overlapping periods. A side benefit is that we can simplify the
handling of concurrency conflict resolution by moving that code into the
record-type-specfic functions; there's no more need to contort the code
layout to keep conflict resolution in front of the RestoreBkpBlocks call.
In connection with that, standardize on zero-based numbering rather than
one-based numbering for referencing the full-page images. In HEAD, I
removed the macros XLR_BKP_BLOCK_1 through XLR_BKP_BLOCK_4. They are
still there in the header files in previous branches, but are no longer
used by the code.
In addition, fix some other bugs identified in the course of making these
changes:
spgRedoAddNode could fail to update the parent downlink at all, if the
parent tuple is in the same page as either the old or new split tuple and
we're not doing a full-page image: it would get fooled by the LSN having
been advanced already. This would result in permanent index corruption,
not just transient failure of concurrent queries.
Also, ginHeapTupleFastInsert's "merge lists" case failed to mark the old
tail page as a candidate for a full-page image; in the worst case this
could result in torn-page corruption.
heap_xlog_freeze() was inconsistent about using a cleanup lock or plain
exclusive lock: it did the former in the normal path but the latter for a
full-page image. A plain exclusive lock seems sufficient, so change to
that.
Also, remove gistRedoPageDeleteRecord(), which has been dead code since
VACUUM FULL was rewritten.
Back-patch to 9.0, where hot standby was introduced. Note however that 9.0
had a significantly different WAL-logging scheme for GIST index updates,
and it doesn't appear possible to make that scheme safe for concurrent hot
standby queries, because it can leave inconsistent states in the index even
between WAL records. Given the lack of complaints from the field, we won't
work too hard on fixing that branch.
This fixes another error in commit 9e8da0f757.
I neglected to make the mark/restore functionality save and restore the
current set of array key values, which led to strange behavior if an
IndexScan with ScalarArrayOpExpr quals was used as the inner side of a
mergejoin. Per bug #7570 from Melese Tesfaye.
The heapam XLog functions are used by other modules, not all of which
are interested in the rest of the heapam API. With this, we let them
get just the XLog stuff in which they are interested and not pollute
them with unrelated includes.
Also, since heapam.h no longer requires xlog.h, many files that do
include heapam.h no longer get xlog.h automatically, including a few
headers. This is useful because heapam.h is getting pulled in by
execnodes.h, which is in turn included by a lot of files.
As noted by Noah Misch, btree_xlog_delete_get_latestRemovedXid is
critically dependent on the assumption that it's examining a consistent
state of the database. This was undocumented though, so the
seemingly-unrelated check for no active HS sessions might be thought to be
merely an optional optimization. Improve comments, and add an explicit
check of reachedConsistency just to be sure.
This function returns InvalidTransactionId (thereby killing all HS
transactions) in several cases that are not nearly unlikely enough for my
taste. This commit doesn't attempt to fix those deficiencies, just
document them.
Back-patch to 9.2, not from any real functional need but just to keep the
branches more closely synced to simplify possible future back-patching.
When we allowed read-only transactions to skip assigning XIDs
we introduced the possibility that a fully deleted btree page
could be reused. This broke the index link sequence which could
then lead to indexscans silently returning fewer rows than would
have been correct. The actual incidence of silent errors from
this is thought to be very low because of the exact workload
required and locking pre-conditions. Fix is to remove pages only
if index page opaque->btpo.xact precedes RecentGlobalXmin.
Noah Misch, reviewed by Simon Riggs
When "vacuuming" a single btree page by removing LP_DEAD tuples, we are not
actually within a vacuum operation, but rather in an ordinary insertion
process that could well be running concurrently with a vacuum. So clearing
the cycleid is incorrect, and could cause the concurrent vacuum to miss
removing tuples that it needs to remove. This is a longstanding bug
introduced by commit e6284649b9 of
2006-07-25. I believe it explains Maxim Boguk's recent report of index
corruption, and probably some other previously unexplained reports.
In 9.0 and up this is a one-line fix; before that we need to introduce a
flag to tell _bt_delitems what to do.
In commit e2c2c2e8b1 I made use of nested
list structures to show which clauses went with which index columns, but
on reflection that's a data structure that only an old-line Lisp hacker
could love. Worse, it adds unnecessary complication to the many places
that don't much care which clauses go with which index columns. Revert
to the previous arrangement of flat lists of clauses, and instead add a
parallel integer list of column numbers. The places that care about the
pairing can chase both lists with forboth(), while the places that don't
care just examine one list the same as before.
The only real downside to this is that there are now two more lists that
need to be passed to amcostestimate functions in case they care about
column matching (which btcostestimate does, so not passing the info is not
an option). Rather than deal with 11-argument amcostestimate functions,
pass just the IndexPath and expect the functions to extract fields from it.
That gets us down to 7 arguments which is better than 11, and it seems
more future-proof against likely additions to the information we keep
about an index path.
The need for this was debated when we put in the index-only-scan feature,
but at the time we had no near-term expectation of having AMs that could
support such scans for only some indexes; so we kept it simple. However,
the SP-GiST AM forces the issue, so let's fix it.
This patch only installs the new API; no behavior actually changes.
This patch creates an API whereby a btree index opclass can optionally
provide non-SQL-callable support functions for sorting. In the initial
patch, we only use this to provide a directly-callable comparator function,
which can be invoked with a bit less overhead than the traditional
SQL-callable comparator. While that should be of value in itself, the real
reason for doing this is to provide a datatype-extensible framework for
more aggressive optimizations, as in Peter Geoghegan's recent work.
Robert Haas and Tom Lane
When wal_level = 'hot_standby' we touched the last page of the
relation during a VACUUM, even if nothing else had happened.
That would alter the LSN of the last block and set the mtime
of the relation file unnecessarily. Noted by Thom Brown.
If we have an inequality key that constrains the other end of the index,
it doesn't directly help us in doing the initial positioning ... but it
does imply a NOT NULL constraint on the index column. If the index stores
nulls at this end, we can use the implied NOT NULL condition for initial
positioning, just as if it had been stated explicitly. This avoids wasting
time when there are a lot of nulls in the column. This is the reverse of
the examples given in bugs #6278 and #6283, which were about failing to
stop early when we encounter nulls at the end of the indexscan.
As pointed out by Naoya Anzai, my previous try at this was a few bricks
shy of a load, because I had forgotten that the initial-positioning logic
might not try to skip over nulls at the end of the index the scan will
start from. We ought to fix that, because it represents an unnecessary
inefficiency, but first let's get the scan-stop logic back to a safe
state. With this patch, we preserve the performance benefit requested
in bug #6278 for the case of scanning forward into NULLs (in a NULLS
LAST index), but the reverse case of scanning backward across NULLs
when there's no suitable initial-positioning qual is still inefficient.
The existing scan-direction-sensitive tests were overly complex, and
failed to stop the scan in cases where it's perfectly legitimate to do so.
Per bug #6278 from Maksym Boguk.
Back-patch to 8.3, which is as far back as the patch applies easily.
Doesn't seem worth sweating over a relatively minor performance issue in
8.2 at this late date. (But note that this was a performance regression
from 8.1 and before, so 8.2 is being left as an outlier.)
In general the data returned by an index-only scan should have the
datatypes originally computed by FormIndexDatum. If the index opclasses
use "storage" datatypes different from their input datatypes, the scan
tuple will not have the same rowtype attributed to the index; but we had
a hard-wired assumption that that was true in nodeIndexonlyscan.c. We'd
already hacked around the issue for the one case where the types are
different in btree indexes (btree name_ops), but this would definitely
come back to bite us if we ever implement index-only scans in GiST.
To fix, require the index AM to explicitly provide the tupdesc for the
tuple it is returning. btree can just pass back the index's tupdesc, but
GiST will have to work harder when and if it supports index-only scans.
I had previously proposed fixing this by allowing the index AM to fill the
scan tuple slot directly; but on reflection that seemed like a module
layering violation, since TupleTableSlots are creatures of the executor.
At least in the btree case, it would also be less efficient, since the
tuple deconstruction work would occur even for rows later found to be
invisible to the scan's snapshot.
Simon Riggs
Heikki Linnakangas
Alvaro Herrera
Tom Lane
Bruce Momjian
Robert Haas
Jeff Davis
Stephen Frost