This covers both regular and inplace changes, since bugs arise at their
intersection. Where marked, these witness extant bugs. Back-patch to
v12 (all supported versions).
Reviewed (in an earlier version) by Robert Haas.
Discussion: https://postgr.es/m/20240512232923.aa.nmisch@google.com
When a pipeline is opened with \startpipeline and not closed, pgbench
will either error on the next transaction with a "already in pipeline
mode" error or successfully end if this was the last transaction --
despite not sending anything that was piped in the pipeline.
Make it an error to reach end of script is reached while there's an
open pipeline.
Backpatch to 14, where pgbench got support for pipelines.
Author: Anthonin Bonnefoy <anthonin.bonnefoy@datadoghq.com>
Reported-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/Za4IObZkDjrO4TcS@paquier.xyz
It draws an unnecessary error in builds compiled without thread support.
Added by commit 038f586d5f, which was backpatched to 14; though in
branch master we no longer support such builds, there's no reason to
have this there, so remove it in all branches since 14.
Reported-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/ZW2G9Ix4nBKLcSSO@paquier.xyz
Starting on 2023-08-03, this intermittently terminated a "pgbench -C"
test in CI. It could affect a high-client-count "pgbench" without "-C".
While parallel reindexdb and vacuumdb reach the same problematic check,
sufficient client count and/or connection turnover is less plausible for
them. Given the lack of examples from the buildfarm or from manual
builds, reproducing this must entail rare operating system
configurations. Also correct the associated error message, which was
wrong for non-Windows. Back-patch to v12, where the pgbench check first
appeared. While v11 vacuumdb has the problematic check, reaching it
with typical vacuumdb usage is implausible.
Reviewed by Thomas Munro.
Discussion: https://postgr.es/m/CA+hUKG+JwvTNdcyJTriy9BbtzF1veSRQ=9M_ZKFn9_LqE7Kp7Q@mail.gmail.com
Failing to do so results in an error when a pgbench script tries to
start a serializable transaction inside a pipeline, because by the time
BEGIN ISOLATION LEVEL SERIALIZABLE is executed, we're already in a
transaction that has acquired a snapshot, so the server rightfully
complains.
We can work around that by preparing all commands in the pipeline before
actually starting the pipeline. This changes the existing code in two
aspects: first, we now prepare each command individually at the point
where that command is about to be executed; previously, we would prepare
all commands in a script as soon as the first command of that script
would be executed. It's hard to see that this would make much of a
difference (particularly since it only affects the first time to execute
each script in a client), but I didn't actually try to measure it.
Secondly, we no longer use PQsendPrepare() in pipeline mode, but only
PQprepare. There's no specific reason for this change other than no
longer needing to do differently in pipeline mode. (Previously we had
no choice, because in pipeline mode PQprepare could not be used.)
Backpatch to 14, where pgbench got support for pipeline mode.
Reported-by: Yugo NAGATA <nagata@sraoss.co.jp>
Discussion: https://postgr.es/m/20210716153013.fc53b1c780b06fccc07a7f0d@sraoss.co.jp
clang 15+ will issue a set-but-not-used warning when the only
use of a variable is in autoincrements (e.g., "foo++;").
That's perfectly sensible, but it detects a few more cases that
we'd not noticed before. Silence the warnings with our usual
methods, such as PG_USED_FOR_ASSERTS_ONLY, or in one case by
actually removing a useless variable.
One thing that we can't nicely get rid of is that with %pure-parser,
Bison emits "yynerrs" as a local variable that falls foul of this
warning. To silence those, I inserted "(void) yynerrs;" in the
top-level productions of affected grammars.
Per recently-established project policy, this is a candidate
for back-patching into out-of-support branches: it suppresses
annoying compiler warnings but changes no behavior. Hence,
back-patch to 9.5, which is as far as these patches go without
issues. (A preliminary check shows that the prior branches
need some other set-but-not-used cleanups too, so I'll leave
them for another day.)
Discussion: https://postgr.es/m/514615.1663615243@sss.pgh.pa.us
Remove meaningless "failures" column from the aggregate logging. It
was just a sum of "serialization failures" and "deadlock failures".
Pointed out by Tom Lane. Patch reviewed by Fabien COELHO.
Discussion: https://postgr.es/m/4183048.1649536705%40sss.pgh.pa.us
Get rid of the separate "FATAL" log level, as it was applied
so inconsistently as to be meaningless. This mostly involves
s/pg_log_fatal/pg_log_error/g.
Create a macro pg_fatal() to handle the common use-case of
pg_log_error() immediately followed by exit(1). Various
modules had already invented either this or equivalent macros;
standardize on pg_fatal() and apply it where possible.
Invent the ability to add "detail" and "hint" messages to a
frontend message, much as we have long had in the backend.
Except where rewording was needed to convert existing coding
to detail/hint style, I have (mostly) resisted the temptation
to change existing message wording.
Patch by me. Design and patch reviewed at various stages by
Robert Haas, Kyotaro Horiguchi, Peter Eisentraut and
Daniel Gustafsson.
Discussion: https://postgr.es/m/1363732.1636496441@sss.pgh.pa.us
Commit 4a39f87acd changed the aggregated log format. Problem is, now
the explanatory paragraph for the log line in the document is too
long. Also the log format included more optional columns, and it's
harder to parse the log lines. This commit tries to solve the
problems.
- There's no optional log columns anymore. If a column is not
meaningful with provided pgbench option, it will be presented as 0.
- Reorder the log columns so that it's easier to parse them.
- Adjust explanatory paragraph for the log line in the doc.
Discussion: https://postgr.es/m/flat/202203280757.3tu4ovs3petm%40alvherre.pgsql
In the wake of commit 4a39f87ac, which noticeably increased the
size of struct StatsData and thereby ParsedScript, Coverity started
to complain that ParsedScript was unreasonably large to be passing
by value. The two places that do this are only used during setup,
so they're not really dragging down benchmark measurements --- but
gratuitous inefficiency is not a good look in a benchmarking program.
Convert to use pointers instead.
Bildfarm member prairiedog reported a pgbench TAP test failure after
commit: 4a39f87acd. This is the second
attempt to fix it. It seems older version of Perl does not accept
"\gN". Replace it with plain old "\N" because actually "\gN" is not
necessary here.
Author: Tatsuo Ishii
Reported-by: Tom Lane
Reviewed-by: Tom Lane, Yugo Nagata
Discussion: https://postgr.es/m/2775989.1648060014%40sss.pgh.pa.us
The Config and Cwd modules were no longer used, but remained imported,
in a number of tests. Remove to keep the imports to the actually used
modules.
Reviewed-by: Andrew Dunstan <andrew@dunslane.net>
Discussion: https://postgr.es/m/A5A074CD-3198-492B-BE5E-7961EFC3733F@yesql.se
Bildfarm member prairiedog reported a pgbench TAP test failure after
commit: 4a39f87acd. This commit attempts
to fix some copy&paste errors introduced in the previous commit.
Author: Yugo Nagata
Reported-by: Tom Lane
Discussion: https://postgr.es/m/2775989.1648060014%40sss.pgh.pa.us
When serialization or deadlock errors are reported by backend, allow
to retry and continue the benchmarking. For this purpose new options
"--max-tries", "--failures-detailed" and "--verbose-errors" are added.
Transactions with serialization errors or deadlock errors will be
repeated after rollbacks until they complete successfully or reach the
maximum number of tries (specified by the --max-tries option), or the
maximum time of tries (specified by the --latency-limit option).
These options can be specified at the same time. It is not possible to
use an unlimited number of tries (--max-tries=0) without the
--latency-limit option or the --time option. By default the option
--max-tries is set to 1, which means transactions with
serialization/deadlock errors are not retried. If the last try fails,
this transaction will be reported as failed, and the client variables
will be set as they were before the first run of this transaction.
Statistics on retries and failures are printed in the progress,
transaction / aggregation logs and in the end with other results (all
and for each script). Also retries and failures are printed
per-command with average latency by using option
(--report-per-command, -r).
Option --failures-detailed prints group failures by basic types
(serialization failures / deadlock failures).
Option --verbose-errors prints distinct reports on errors and failures
(errors without retrying) by type with detailed information like which
limit for retries was violated and how far it was exceeded for the
serialization/deadlock failures.
Patch originally written by Marina Polyakova then Yugo Nagata
inherited the discussion and heavily modified the patch to make it
commitable.
Authors: Yugo Nagata, Marina Polyakova
Reviewed-by: Fabien Coelho, Tatsuo Ishii, Alvaro Herrera, Kevin Grittner, Andres Freund, Arthur Zakirov, Alexander Korotkov, Teodor Sigaev, Ildus Kurbangaliev
Discussion: https://postgr.es/m/flat/72a0d590d6ba06f242d75c2e641820ec%40postgrespro.ru
Move scanint8() to numutils.c and rename to pg_strtoint64(). We
already have a "16" and "32" version of that, and the code inside the
functions was aligned, so this move makes all three versions
consistent. The API is also changed to no longer provide the errorOK
case. Users that need the error checking can use strtoi64().
Reviewed-by: John Naylor <john.naylor@enterprisedb.com>
Discussion: https://www.postgresql.org/message-id/flat/b239564c-cad0-b23e-c57e-166d883cb97d@enterprisedb.com
Standardize on xoroshiro128** as our basic PRNG algorithm, eliminating
a bunch of platform dependencies as well as fundamentally-obsolete PRNG
code. In addition, this API replacement will ease replacing the
algorithm again in future, should that become necessary.
xoroshiro128** is a few percent slower than the drand48 family,
but it can produce full-width 64-bit random values not only 48-bit,
and it should be much more trustworthy. It's likely to be noticeably
faster than the platform's random(), depending on which platform you
are thinking about; and we can have non-global state vectors easily,
unlike with random(). It is not cryptographically strong, but neither
are the functions it replaces.
Fabien Coelho, reviewed by Dean Rasheed, Aleksander Alekseev, and myself
Discussion: https://postgr.es/m/alpine.DEB.2.22.394.2105241211230.165418@pseudo
Previously failures of initial connection and logfile open caused pgbench
to proceed the benchmarking, report the incomplete results and exit with
status 2. It didn't make sense to proceed the benchmarking even when
pgbench could not start as prescribed.
This commit improves pgbench so that early errors that occur when
starting benchmark such as those failures should make pgbench exit
immediately with status 1.
Author: Yugo Nagata
Reviewed-by: Fabien COELHO, Kyotaro Horiguchi, Fujii Masao
Discussion: https://postgr.es/m/TYCPR01MB5870057375ACA8A73099C649F5349@TYCPR01MB5870.jpnprd01.prod.outlook.com
The five modules in our TAP test framework all had names in the top
level namespace. This is unwise because, even though we're not
exporting them to CPAN, the names can leak, for example if they are
exported by the RPM build process. We therefore move the modules to the
PostgreSQL::Test namespace. In the process PostgresNode is renamed to
Cluster, and TestLib is renamed to Utils. PostgresVersion becomes simply
PostgreSQL::Version, to avoid possible confusion about what it's the
version of.
Discussion: https://postgr.es/m/aede93a4-7d92-ef26-398f-5094944c2504@dunslane.net
Reviewed by Erik Rijkers and Michael Paquier
CIC and REINDEX CONCURRENTLY assume backends see their catalog changes
no later than each backend's next transaction start. That failed to
hold when a backend absorbed a relevant invalidation in the middle of
running RelationBuildDesc() on the CIC index. Queries that use the
resulting index can silently fail to find rows. Fix this for future
index builds by making RelationBuildDesc() loop until it finishes
without accepting a relevant invalidation. It may be necessary to
reindex to recover from past occurrences; REINDEX CONCURRENTLY suffices.
Back-patch to 9.6 (all supported versions).
Noah Misch and Andrey Borodin, reviewed (in earlier versions) by Andres
Freund.
Discussion: https://postgr.es/m/20210730022548.GA1940096@gust.leadboat.com
Previously socket errors such as invalid socket or socket wait method failures
during benchmark caused pgbench to exit with status 0. Instead, errors during
the run should result in exit status 2.
Back-patch to v12 where pgbench started reporting exit status.
Original complaint and patch by Hayato Kuroda.
Author: Yugo Nagata, Fabien COELHO
Reviewed-by: Kyotaro Horiguchi, Fujii Masao
Discussion: https://postgr.es/m/TYCPR01MB5870057375ACA8A73099C649F5349@TYCPR01MB5870.jpnprd01.prod.outlook.com
The failure of socket wait method like "select()" doesn't terminate pgbench.
So the log level of error message when that failure happens should be ERROR.
But previously FATAL was used in that case.
Back-patch to v13 where pgbench started using common logging API.
Author: Yugo Nagata, Fabien COELHO
Reviewed-by: Kyotaro Horiguchi, Fujii Masao
Discussion: https://postgr.es/m/20210617005934.8bd37bf72efd5f1b38e6f482@sraoss.co.jp
When throttling is used, transactions that lag behind schedule by
more than the latency limit are counted and reported as skipped.
Previously, there was the case where pgbench counted all skipped
transactions even if the timer specified in -T option was exceeded.
This could take a very long time to do that especially when unrealistically
high rate setting in -R option caused quite a lot of transactions that
lagged behind schedule. This could prevent pgbench from ending
immediately, and so pgbench could look like it got stuck to users.
To fix the issue, this commit changes pgbench so that it stops counting
skipped transactions as soon as the timer is exceeded. The timer can
make pgbench end soon even when there are lots of skipped transactions
that have not been counted yet.
Note that there is no guarantee that all skipped transactions are
counted under -T though there is under -t. This is OK in practice
because it's very unlikely to happen with realistic setting. Also this is
not the issue that this commit newly introdues. There used to be
the case where pgbench ended without counting all skipped
transactions since before.
Back-patch to v14. Per discussion, we decided not to bother
back-patch to the stable branches because it's hard to imagine
the issue happens in practice (with realistic setting).
Author: Yugo Nagata, Fabien COELHO
Reviewed-by: Greg Sabino Mullane, Fujii Masao
Discussion: https://postgr.es/m/20210613040151.265ff59d832f835bbcf8b3ba@sraoss.co.jp
While populating the pgbench_accounts table, plain COPY was
unconditionally used. By changing it to COPY FREEZE, the time for
VACUUM is significantly reduced, thus the total time of "pgbench -i"
is also reduced. This only happens if pgbench runs against PostgreSQL
14 or later because COPY FREEZE in previous versions of PostgreSQL does
not bring the benefit. Also if partitioning is used, COPY FREEZE
cannot be used. In this case plain COPY will be used too.
Author: Tatsuo Ishii
Discussion: https://postgr.es/m/20210308.143907.2014279678657453983.t-ishii@gmail.com
Reviewed-by: Fabien COELHO, Laurenz Albe, Peter Geoghegan, Dean Rasheed
When -C/--connect option is specified, pgbench establishes and closes
the connection for each transaction. In this case pgbench needs to
measure the times taken for all those connections and disconnections,
to include the average connection time in the benchmark result.
But previously pgbench could not measure those disconnection delays.
To fix the bug, this commit makes pgbench measure the disconnection
delays whenever the connection is closed at the end of transaction,
if -C/--connect option is specified.
Back-patch to v14. Per discussion, we concluded not to back-patch to v13
or before because changing that behavior in stable branches would
surprise users rather than providing benefits.
Author: Yugo Nagata
Reviewed-by: Fabien COELHO, Tatsuo Ishii, Asif Rehman, Fujii Masao
Discussion: https://postgr.es/m/20210614151155.a393bc7d8fed183e38c9f52a@sraoss.co.jp
Commit 547f04e734 changed pgbench so that it used the measurement result
of connection delays in its benchmark report only when -C/--connect option
is specified. But previously those delays were unnecessarily measured
even when that option is not specified. Which was a waste of cycles.
This commit improves pgbench so that it avoids such unnecessary measurement.
Back-patch to v14 where commit 547f04e734 first appeared.
Author: Yugo Nagata
Reviewed-by: Fabien COELHO, Asif Rehman, Fujii Masao
Discussion: https://postgr.es/m/20210614151155.a393bc7d8fed183e38c9f52a@sraoss.co.jp
Up to now we did a PQconsumeInput() for each pipelined query, asking the OS
for more input - which it often won't have, as all results might already have
been sent. That turns out to have a noticeable performance impact.
Alvaro Herrera reviewed the idea to add the PQisBusy() check, but not this
concrete patch.
Author: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/20210720180039.23rivhdft3l4mayn@alap3.anarazel.de
Backpatch: 14, where libpq/pgbench pipelining was introduced.
There is only one constructor now for PostgresNode, with the idiomatic
name 'new'. The method is not exported by the class, and must be called
as "PostgresNode->new('name',[args])". All the TAP tests that use
PostgresNode are modified accordingly. Third party scripts will need
adjusting, which is a fairly mechanical process (I just used a sed
script).
The following changes are done:
- In pg_archivecleanup, the cleanup of older WAL segments would never
fail immediately.
- In pgbench, the initialization of a thread barrier would not fail
hard.
- In pg_recvlogical, a stat() failure never got the call.
- In pg_basebackup, two chmod() reported a failure without exit()'ing
when unpacking some tar data freshly received. It may be possible to
continue writing some data even after this failure, but that could be
confusing to the user at the end.
These are arguably bugs, but they would happen for code paths where a
failure is unlikely going to happen, so no backpatch is done.
Reviewed-by: Robert Haas, Fabien Coelho
Discussion: https://postgr.es/m/YQDMdB+B68yePFeT@paquier.xyz
Most of the integer options for command-line binaries now make use of a
single routine able to do the job, fixing issues with the detection of
sloppy values caused for example by the use of atoi(), that fails on
strings beginning with numerical characters with junk trailing
characters.
This commit cuts down the number of strings requiring translation by 26
per my count, switching the code to have two error types for invalid and
out-of-range values instead.
Much more could be done here, with float or even int64 options, but
int32 was the most appealing case as it is possible to rely on strtol()
to do the job reliably. Note that there are some exceptions for now,
like pg_ctl or pg_upgrade that use their own logging logic. A couple of
negative TAP tests required some adjustments for the new errors
generated.
pg_dump and pg_restore tracked the maximum number of parallel jobs
within the option parsing. The code is refactored a bit to track that
in the code dedicated to parallelism instead.
Author: Kyotaro Horiguchi, Michael Paquier
Reviewed-by: David Rowley, Álvaro Herrera
Discussion: https://postgr.es/m/CALj2ACXqdG9WhqVoJ9zYf-iZt7sgK7Szv5USs=he6NnWQ2ofTA@mail.gmail.com
Commit 547f04e changed pgbench to use microsecond accounting, but
introduced a couple of logging and aggregation bugs:
1. We print Unix epoch timestamps so that you can correlate them with
other logs, but these were inadvertently changed to use a
system-dependent reference epoch. Compute Unix timestamps, and begin
aggregation intervals on the boundaries of whole Unix epoch seconds, as
before.
2. The user-supplied aggregation interval needed to be scaled.
Back-patch to 14.
Author: Fabien COELHO <coelho@cri.ensmp.fr>
Author: Yugo NAGATA <nagata@sraoss.co.jp>
Author: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
Reported-by: YoungHwan Joo <rulyox@gmail.com>
Reported-by: Gregory Smith <gregsmithpgsql@gmail.com>
Discussion: https://postgr.es/m/CAF7igB1r6wRfSCEAB-iZBKxkowWY6%2BdFF2jObSdd9%2BiVK%2BvHJg%40mail.gmail.com
Discussion: https://postgr.es/m/CAHLJuCW_8Vpcr0%3Dt6O_gozrg3wXXWXZXDioYJd3NhvKriqgpfQ%40mail.gmail.com
Instead use the common/int.h functions to check for integer overflow
in a more C-standard-compliant fashion. This is motivated by recent
failures on buildfarm member moonjelly, where it appears that
development-tip gcc is optimizing without regard to the -fwrapv
switch. Presumably that's a gcc bug that will be fixed soon, but
we might as well install cleaner coding here rather than wait.
(This does not address the question of whether we'll ever be able
to get rid of using -fwrapv. Testing shows that this spot is the
only place where doing so creates visible regression test failures,
but unfortunately that proves very little.)
Back-patch to v12. The common/int.h functions exist in v11, but
that branch doesn't use them in any client-side code. I judge
that this case isn't interesting enough in the real world to take
even a small risk of issues from being the first such use.
Tom Lane and Fabien Coelho
Discussion: https://postgr.es/m/73927.1624815543@sss.pgh.pa.us
002_pgbench_no_server was printing some array pointers instead of the
actual contents of those arrays for the expected outputs of stdout and
stderr for a tested command. This does not add any new information that
can help with debugging as the test names allow to track failure
locations, if any.
This commit simply removes those logs as the rest of the printed
information is redundant with command_checks_all().
Per discussion with Andrew Dunstan and Álvaro Herrera.
Discussion: https://postgr.es/m/YNXNFaG7IgkzZanD@paquier.xyz
Backpatch-through: 11
This fixes a couple of problems within the so-said code of this commit
subject:
- Replace the use of open() with slurp_file(), fixing an issue reported
by buildfarm member fairywren whose perl installation keep around CRLF
characters, causing the matching patterns for the logs to fail.
- Remove the eval block, which is not really necessary.
This set of issues has come into light after fixing a different issue
with c13585fe, and this is wrong since this code has been introduced.
Reported-by: Andrew Dunstan, and buildfarm member fairywren
Author: Michael Paquier
Reviewed-by: Andrew Dunstan
Discussion: https://postgr.es/m/0f49303e-7784-b3ee-200b-cbf67be2eb9e@dunslane.net
Backpatch-through: 11
fairywren is not happy with the pattern checks introduced by c13585f.
I am not sure if this outlines a bug in pgbench or if the regex patterns
used in the tests are too restrictive for this buildfarm member's
environment. This adds more debugging information to show the log
entries that do not match with the expected pattern, to help in finding
out what's happening. That seems like a good addition in the long-term
anyway as that may not be the only issue in this area.
Discussion: https://postgr.es/m/YNUad2HvgW+6eXyo@paquier.xyz
The logic checking for the format of per-thread logs used grep() with
directly "$re", which would cause the test to consider all the logs as
a match without caring about their format at all. Using "/$re/" makes
grep() perform a regex test, which is what we want here.
While on it, improve some of the tests to be more picky with the
patterns expected and add more comments to describe the tests.
Issue discovered while digging into a separate patch.
Author: Fabien Coelho, Michael Paquier
Discussion: https://postgr.es/m/YNPsPAUoVDCpPOGk@paquier.xyz
Backpatch-through: 11
