These commands allow assignment of values produced by queries to pgbench
variables, where they can be used by further commands. \gset terminates
a command sequence (just like a bare semicolon); \cset separates
multiple queries in a compound command, like an escaped semicolon (\;).
A prefix can be provided to the \-command and is prepended to the name
of each output column to produce the final variable name.
This feature allows pgbench scripts to react meaningfully to the actual
database contents, allowing more powerful benchmarks to be written.
Authors: Fabien Coelho, Álvaro Herrera
Reviewed-by: Amit Langote <Langote_Amit_f8@lab.ntt.co.jp>
Reviewed-by: Stephen Frost <sfrost@snowman.net>
Reviewed-by: Pavel Stehule <pavel.stehule@gmail.com>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Tatsuo Ishii <ishii@sraoss.co.jp>
Reviewed-by: Rafia Sabih <rafia.sabih@enterprisedb.com>
Discussion: https://postgr.es/m/alpine.DEB.2.20.1607091005330.3412@sto
This removes a portion of infrastructure introduced by fe0a0b5 to allow
compilation of Postgres in environments where no strong random source is
available, meaning that there is no linking to OpenSSL and no
/dev/urandom (Windows having its own CryptoAPI). No systems shipped
this century lack /dev/urandom, and the buildfarm is actually not
testing this switch at all, so just remove it. This simplifies
particularly some backend code which included a fallback implementation
using shared memory, and removes a set of alternate regression output
files from pgcrypto.
Author: Michael Paquier
Reviewed-by: Tom Lane
Discussion: https://postgr.es/m/20181230063219.GG608@paquier.xyz
This commit continues the code improvements started by commit
12788ae49e. With this commit, state machine transitions are better
contained in the routine that was called doCustom() and is now called
advanceConnectionState -- the resulting code is easier to reason about,
since there are no state changes occuring in the outer layer.
This change is prompted by future patches to add more features to
pgbench, which will need to effect some more surgery to this code.
Fabien's original had all the machine state changes inside one routine,
but I (Álvaro) thought that a subroutine to handle command execution is
more straightforward to review, so this commit does not match Fabien's
submission closely. If something is broken, it's probably my fault.
Author: Fabien Coelho, Álvaro Herrera
Reviewed-by: Kirk Jamison
Discussion: https://postgr.es/m/alpine.DEB.2.21.1808111104320.1705@lancre
This becomes useful when used to retry a transaction after a
serialization error or deadlock abort. (We don't yet have that feature,
but this is preparation for it.)
While at it, use separate random state for thread administratrivia such
as deciding which script to run, how long to delay for throttling, or
whether to log a message when sampling; this not only makes these tasks
independent of each other, but makes the actual thread run
deterministic.
Author: Marina Polyakova
Reviewed-by: Fabien Coelho
Discussion: https://postgr.es/m/72a0d590d6ba06f242d75c2e641820ec@postgrespro.ru
When an error occurs during a benchmark run, exit with a nonzero exit
code and write a message at the end. Previously, it would just print
the error message when it happened but then proceed to print the run
summary normally and exit with status 0. To still allow
distinguishing setup from run-time errors, we use exit status 2 for
the new state, whereas existing errors during pgbench initialization
use exit status 1.
Reviewed-by: Fabien COELHO <coelho@cri.ensmp.fr>
It's fairly silly to truncate the throttle_delay to integer when the only
math we ever do with it requires converting back to double. Furthermore,
given that people are starting to complain about restrictions like only
supporting 1K client connections, I don't think we're very far away from
situations where the precision loss matters. As the code stood, for
example, there's no difference between --rate 100001 and --rate 111111;
both get converted to throttle_delay = 9. Somebody trying to run 100
threads and have each one dispatch around 1K TPS would find this lack of
precision rather surprising, especially since the required per-thread
delays are around 1ms, well within the timing precision of modern systems.
Previously, pgbench always used select(2) for this purpose, but that's
problematic for very high client counts, because select() can't deal
with file descriptor numbers larger than FD_SETSIZE. It's pretty common
for that to be only 1024 or so, whereas modern OSes can allow many more
open files than that. Using poll(2) would surmount that problem, but it
creates another one: poll()'s timeout resolution is only 1ms, which is
poor enough to cause problems with --rate specifications approaching or
exceeding 1K TPS.
On platforms that have ppoll(2), which includes Linux and recent
FreeBSD, we can use that to avoid the FD_SETSIZE problem without any
loss of timeout resolution. Hence, add configure logic to test for
ppoll(), and use it if available.
This patch introduces an abstraction layer into pgbench that could
be extended to support other kernel event-wait APIs such as kevents.
But actually adding such support is a matter for some future patch.
Doug Rady, reviewed by Robert Haas and Fabien Coelho, and whacked around
a good bit more by me
Discussion: https://postgr.es/m/23D017C9-81B7-484D-8490-FD94DEC4DF59@amazon.com
Update links that resulted in redirects. Most are changes from http to
https, but there are also some other minor edits. (There are still some
redirects where the target URL looks less elegant than the one we
currently have. I have left those as is.)
Recent gcc can warn about switch-case fall throughs that are not
explicitly labeled as intentional. This seems like a good thing,
so clean up the warnings exposed thereby by labeling all such
cases with comments that gcc will recognize.
In files that already had one or more suitable comments, I generally
matched the existing style of those. Otherwise I went with
/* FALLTHROUGH */, which is one of the spellings approved at the
more-restrictive-than-default level -Wimplicit-fallthrough=4.
(At the default level you can also spell it /* FALL ?THRU */,
and it's not picky about case. What you can't do is include
additional text in the same comment, so some existing comments
containing versions of this aren't good enough.)
Testing with gcc 8.0.1 (Fedora 28's current version), I found that
I also had to put explicit "break"s after elog(ERROR) or ereport(ERROR);
apparently, for this purpose gcc doesn't recognize that those don't
return. That seems like possibly a gcc bug, but it's fine because
in most places we did that anyway; so this amounts to a visit from the
style police.
Discussion: https://postgr.es/m/15083.1525207729@sss.pgh.pa.us
Compilation failed for lack of an #ifdef on builds without
pg_strong_random(). Also fix relevant error messages to meet
project style guidelines.
Fabien Coelho, further adjusted by me
Discussion: https://postgr.es/m/32390.1522464534@sss.pgh.pa.us
Setting random could increase reproducibility of test in some cases. Patch
suggests three providers for seed: time (default), strong random
generator (if available) and unsigned constant. Seed could be set from
command line or enviroment variable.
Author: Fabien Coelho
Reviewed by: Chapman Flack
Discussion: https://www.postgresql.org/message-id/flat/20160407082711.q7iq3ykffqxcszkv@alap3.anarazel.de
Fix the warnings created by the compiler warning options
-Wformat-overflow=2 -Wformat-truncation=2, supported since GCC 7. This
is a more aggressive variant of the fixes in
6275f5d28a, which GCC 7 warned about by
default.
The issues are all harmless, but some dubious coding patterns are
cleaned up.
One issue that is of external interest is that BGW_MAXLEN is increased
from 64 to 96. Apparently, the old value would cause the bgw_name of
logical replication workers to be truncated in some circumstances.
But this doesn't actually add those warning options. It appears that
the warnings depend a bit on compilation and optimization options, so it
would be annoying to have to keep up with that. This is more of a
once-in-a-while cleanup.
Reviewed-by: Michael Paquier <michael@paquier.xyz>
All of these are false positives, but in each case a fair amount of
analysis is needed to see that, and it's not too surprising that not all
compilers are smart enough. (In particular, in the logtape.c case, a
compiler lacking the knowledge provided by the Assert would almost surely
complain, so that this warning will be seen in any non-assert build.)
Some of these are of long standing while others are pretty recent,
but it only seems worth fixing them in HEAD.
Jaime Casanova, tweaked a bit by me
Discussion: https://postgr.es/m/CAJGNTeMcYAMJdPAom52dppLMtF-UnEZi0dooj==75OEv1EoBZA@mail.gmail.com
Added:
- variable now might contain integer, double, boolean and null values
- functions ln, exp
- logical AND/OR/NOT
- bitwise AND/OR/NOT/XOR
- bit right/left shift
- comparison operators
- IS [NOT] (NULL|TRUE|FALSE)
- conditional choice (in form of when/case/then)
New operations and functions allow to implement more complicated test scenario.
Author: Fabien Coelho with minor editorization by me
Reviewed-By: Pavel Stehule, Jeevan Ladhe, me
Discussion: https://www.postgresql.org/message-id/flat/alpine.DEB.2.10.1604030742390.31618@sto
pgbench can skip some transactions when both -R and -L options are used.
Previously, this resulted in slightly silly statistics both in progress
reports and final output, because the skipped transactions were counted
as executed for TPS and related stats. Discount skipped xacts in TPS
numbers, and also when figuring the percentage of xacts exceeding the
latency limit.
Also, don't print per-script skipped-transaction counts when there is
only one script. That's redundant with the overall count, and it's
inconsistent with the fact that we don't print other per-script stats
when there's only one script. Clean up some unnecessary interactions
between what should be independent options that were due to that
decision.
While at it, avoid division-by-zero in cases where no transactions were
executed. While on modern platforms this would generally result in
printing "NaN" rather than a crash, that isn't spelled consistently
across platforms and it would confuse many people. Skip the relevant
output entirely when practical, else print zeroes.
Fabien Coelho, reviewed by Steve Singer, additional hacking by me
Discussion: https://postgr.es/m/26654.1505232433@sss.pgh.pa.us
This feature caters to specialized use-cases such as running the normal
pgbench scenario with nonstandard indexes, or inserting other actions
between steps of the initialization sequence. The normal sequence of
initialization actions is broken down into half a dozen steps which can
be executed in a user-specified order, to the extent to which that's
sensible. The actions themselves aren't changed, except to make them
more robust against nonstandard uses:
* all four tables are now dropped in one DROP command, to reduce
assumptions about what foreign key relationships exist;
* all four tables are now truncated at the start of the data load
step, for consistency;
* the foreign key creation commands now specify constraint names, to
prevent accidentally creating duplicate constraints by executing the
'f' step twice.
Make some cosmetic adjustments in the messages emitted by pgbench
so that it's clear which steps are getting run, and so that the
messages agree with the documented names of the steps.
In passing, fix failure to enforce that the -v option is used only
in benchmarking mode.
Masahiko Sawada, reviewed by Fabien Coelho, editorialized a bit by me
Discussion: https://postgr.es/m/CAD21AoCsz0ZzfCFcxYZ+PUdpkDd5VsCSG0Pre_-K1EgokCDFYA@mail.gmail.com
If --rate was used to throttle pgbench, it failed to sleep when it had
nothing to do, leading to a busy-wait with 100% CPU usage. This bug was
introduced in the refactoring in v10. Before that, sleep() was called with
a timeout, even when there were no file descriptors to wait for.
Reported by Jeff Janes, patch by Fabien COELHO. Backpatch to v10.
Discussion: https://www.postgresql.org/message-id/CAMkU%3D1x5hoX0pLLKPRnXCy0T8uHoDvXdq%2B7kAM9eoC9_z72ucw%40mail.gmail.com
process_backslash_command would drop the last character of the input
command on the assumption that it was a newline. Given a non newline
terminated input file, this could result in dropping the last character
of the command. Fix that by doing an actual test that we're removing
a newline.
While at it, allow for Windows newlines (\r\n), and suppress multiple
newlines if any. I do not think either of those cases really occur,
since (a) we read script files in text mode and (b) the lexer stops
when it hits a newline. But it's cheap enough and it provides a
stronger guarantee about what the result string looks like.
This is just cosmetic, I think, since the possibly-overly-chomped
line was only used for display not for further processing. So
it doesn't seem necessary to back-patch.
Fabien Coelho, reviewed by Nikolay Shaplov, whacked around a bit by me
Discussion: https://postgr.es/m/alpine.DEB.2.20.1704171422500.4025@lancre
With --latency-limit, transactions might get skipped even beyond the
transaction count limit specified by -t, throwing off the expected
number of transactions and thus the denominator for later stats.
Be sure to stop skipping transactions once -t is reached.
Also, include skipped transactions in the "cnt" fields; this requires
discounting them again in a couple of places, but most places are
better off with this definition. In particular this is needed to
get correct overall stats for the combination of -R/-L/-t options.
Merge some more processing into processXactStats() to simplify this.
In passing, add a check that --progress-timestamp is specified only
when --progress is.
We might consider back-patching this, but given that it only matters
for a combination of options, and given the lack of field complaints,
consensus seems to be not to bother.
Fabien Coelho, reviewed by Nikolay Shaplov
Discussion: https://postgr.es/m/alpine.DEB.2.20.1704171422500.4025@lancre
Previously the -M switch had to appear before any switch that directly
or indirectly specified a benchmarking script. This was both confusing
and inadequately documented, as per gripe from Tatsuo Ishii. We can
remove the restriction at the cost of making an extra pass over the
lists of SQL commands, which seems like a cheap price (the string scans
themselves likely cost much more). The change is just to not extract
parameters from the SQL commands until we have finished parsing the
switches and know the final value of -M.
Per discussion, we'll treat this as a low-grade bug fix and sneak it
into v10, rather than holding it for v11.
Tom Lane, reviewed by Tatsuo Ishii and Fabien Coelho
Discussion: https://postgr.es/m/20170802.110328.1963639094551443169.t-ishii@sraoss.co.jp
Discussion: https://postgr.es/m/10208.1502465077@sss.pgh.pa.us
Don't move parenthesized lines to the left, even if that means they
flow past the right margin.
By default, BSD indent lines up statement continuation lines that are
within parentheses so that they start just to the right of the preceding
left parenthesis. However, traditionally, if that resulted in the
continuation line extending to the right of the desired right margin,
then indent would push it left just far enough to not overrun the margin,
if it could do so without making the continuation line start to the left of
the current statement indent. That makes for a weird mix of indentations
unless one has been completely rigid about never violating the 80-column
limit.
This behavior has been pretty universally panned by Postgres developers.
Hence, disable it with indent's new -lpl switch, so that parenthesized
lines are always lined up with the preceding left paren.
This patch is much less interesting than the first round of indent
changes, but also bulkier, so I thought it best to separate the effects.
Discussion: https://postgr.es/m/E1dAmxK-0006EE-1r@gemulon.postgresql.org
Discussion: https://postgr.es/m/30527.1495162840@sss.pgh.pa.us
Change pg_bsd_indent to follow upstream rules for placement of comments
to the right of code, and remove pgindent hack that caused comments
following #endif to not obey the general rule.
Commit e3860ffa4d wasn't actually using
the published version of pg_bsd_indent, but a hacked-up version that
tried to minimize the amount of movement of comments to the right of
code. The situation of interest is where such a comment has to be
moved to the right of its default placement at column 33 because there's
code there. BSD indent has always moved right in units of tab stops
in such cases --- but in the previous incarnation, indent was working
in 8-space tab stops, while now it knows we use 4-space tabs. So the
net result is that in about half the cases, such comments are placed
one tab stop left of before. This is better all around: it leaves
more room on the line for comment text, and it means that in such
cases the comment uniformly starts at the next 4-space tab stop after
the code, rather than sometimes one and sometimes two tabs after.
Also, ensure that comments following #endif are indented the same
as comments following other preprocessor commands such as #else.
That inconsistency turns out to have been self-inflicted damage
from a poorly-thought-through post-indent "fixup" in pgindent.
This patch is much less interesting than the first round of indent
changes, but also bulkier, so I thought it best to separate the effects.
Discussion: https://postgr.es/m/E1dAmxK-0006EE-1r@gemulon.postgresql.org
Discussion: https://postgr.es/m/30527.1495162840@sss.pgh.pa.us
The new indent version includes numerous fixes thanks to Piotr Stefaniak.
The main changes visible in this commit are:
* Nicer formatting of function-pointer declarations.
* No longer unexpectedly removes spaces in expressions using casts,
sizeof, or offsetof.
* No longer wants to add a space in "struct structname *varname", as
well as some similar cases for const- or volatile-qualified pointers.
* Declarations using PG_USED_FOR_ASSERTS_ONLY are formatted more nicely.
* Fixes bug where comments following declarations were sometimes placed
with no space separating them from the code.
* Fixes some odd decisions for comments following case labels.
* Fixes some cases where comments following code were indented to less
than the expected column 33.
On the less good side, it now tends to put more whitespace around typedef
names that are not listed in typedefs.list. This might encourage us to
put more effort into typedef name collection; it's not really a bug in
indent itself.
There are more changes coming after this round, having to do with comment
indentation and alignment of lines appearing within parentheses. I wanted
to limit the size of the diffs to something that could be reviewed without
one's eyes completely glazing over, so it seemed better to split up the
changes as much as practical.
Discussion: https://postgr.es/m/E1dAmxK-0006EE-1r@gemulon.postgresql.org
Discussion: https://postgr.es/m/30527.1495162840@sss.pgh.pa.us
As a consequence of commit 1d63f7d2d, on platforms with CLOCK_MONOTONIC,
you got some random timescale or other instead of standard Unix timestamps
as expected. I'd attempted to fix pgbench for that change in commits
74baa1e3b and 67a875355, but missed this place. Fix in the same way as
those previous commits, ie, just eat the cost of an extra gettimeofday();
one extra syscall per progress report isn't worth sweating over. Per
report from Jeff Janes.
In passing, use snprintf not sprintf for this purpose. I don't think
there's any chance of actual buffer overrun, but it just looks safer.
Discussion: https://postgr.es/m/CAMkU=1zrQaPwBN+NcBd3pWCb=vWaiL=mmWfJjDJjh-a7eVr-Og@mail.gmail.com
Not only did it not accept --builtin as a synonym for -b, but what it did
accept as a synonym was --tpc-b (huh?), which it got even further wrong
by marking as no_argument, so that if you did try that you got a core
dump. I suppose this is leftover from some early design for the new
switches added by commit 8bea3d221, but it's still pretty sloppy work.
Per bug #14580 from Stepan Pesternikov. Back-patch to 9.6 where the
error was introduced.
Report: https://postgr.es/m/20170307123347.25054.73207@wrigleys.postgresql.org
For aggregated logging, pg_bench supposed that printing the integer part of
INSTR_TIME_GET_DOUBLE() would produce a Unix timestamp. That was already
broken on Windows, and it's about to get broken on most other platforms as
well. As in commit 74baa1e3b, we can remove the entanglement at the price
of one extra syscall per transaction; though here it seems more convenient
to use time(NULL) instead of gettimeofday(), since we only need
integral-second precision.
I took the time to do some wordsmithing on the documentation about
pgbench's logging features, too.
Discussion: https://postgr.es/m/8837.1483216839@sss.pgh.pa.us
This code was presuming undue familiarity with the contents of the
instr_time struct. That was already broken on Windows, and it's about
to get broken on most other platforms as well. The simplest solution
that preserves the current output definition is to just do our own
gettimeofday() call here. Realistically, the extra cost is probably
negligible in comparison to everything else that's going on in a
pgbench transaction, so it's not worth sweating over.
On Windows, the precision delivered by gettimeofday() is lower than
one could wish, but this is still a big improvement over printing
zeroes, as the code did before.
Discussion: https://postgr.es/m/8837.1483216839@sss.pgh.pa.us
Commit 41124a91e6 allowed the
transaction log file prefix to be changed but left in place the
existing 64-character limit on the total length of a log file name.
It's possible that could be inconvenient for somebody, so increase the
limit to MAXPGPATH, which ought to be enough for anybody.
Per a suggestion from Tom Lane.
The result of FD_ISSET() doesn't necessarily fit in a bool, though
assigning it to one might accidentally work depending on platform and which
socket FD number is being inquired of. Rewrite to test it with if(),
rather than making any specific assumption about the result width,
to match the way every other such call in PG is written.
Don't break out of the input_mask-filling loop after finding the first
client that we're waiting for results from. That mostly breaks parallel
query management.
Also, if we choose not to call select(), be sure to clear out any bits
the mask-filling loop might have set, so that we don't accidentally call
doCustom for clients we don't know have input. Doing so would likely
be harmless, but it's a waste of cycles and doesn't seem to be intended.
Make this_usec wide enough. (Yeah, the value would usually fit in an
int, but then why are we using int64 everywhere else?)
Minor cosmetic adjustments, mostly comment improvements.
Problems introduced by commit 12788ae49. The first issue was discovered
by buildfarm testing, the others by code review.
The doCustom() function had grown into quite a mess. Rewrite it, in a more
explicit state machine style, for readability.
This also fixes one minor bug: if a script consisted entirely of meta
commands, doCustom() never returned to the caller, so progress reports
with the -P option were not printed. I don't want to backpatch this
refactoring, and the bug is quite insignificant, so only commit this to
master, and leave the bug unfixed in back-branches.
Review and original bug report by Fabien Coelho.
Discussion: <alpine.DEB.2.20.1607090850120.3412@sto>
The way "latency average" was printed was differently if it was calculated
from the overall run time or was measured on a per-transaction basis.
Also, the per-script weight is a test parameter, rather than a result, so
use the "weight: %f" style for that.
Backpatch to 9.6, since the inconsistency on "latency average" was
introduced there.
Fabien Coelho
Discussion: <alpine.DEB.2.20.1607131015370.7486@sto>
