Several users of extensions complained of crashes in parallel workers
that turned out to be due to syscache access from their _PG_init()
functions. Reorder the initialization of parallel workers so that
libraries are restored after the caches are initialized, and inside a
transaction.
This was reported in bug #15350 and elsewhere. We don't consider it
to be a bug: extensions shouldn't do that, because then they can't be
used in shared_preload_libraries. However, it's a fairly obscure
hazard and these extensions worked in practice before parallel query
came along. So let's make it work. Later commits might add a warning
message and eventually an error.
Back-patch to 9.6, where parallel query landed.
Author: Thomas Munro
Reviewed-by: Amit Kapila
Reported-by: Kieran McCusker, Jimmy
Discussion: https://postgr.es/m/153512195228.1489.8545997741965926448%40wrigleys.postgresql.org
Commit 9da0cc3528, which introduced parallel CREATE INDEX, failed to
propagate relmapper.c backend local cache state to parallel worker
processes. This could result in parallel index builds against mapped
catalog relations where the leader process (participating as a worker)
scans the new, pristine relfilenode, while worker processes scan the
obsolescent relfilenode. When this happened, the final index structure
was typically not consistent with the owning table's structure. The
final index structure could contain entries formed from both heap
relfilenodes. Only rebuilds on mapped catalog relations that occur as
part of a VACUUM FULL or CLUSTER could become corrupt in practice, since
their mapped relation relfilenode swap is what allows the inconsistency
to arise.
On master, fix the problem by propagating the required relmapper.c
backend state as part of standard parallel initialization (Cf. commit
29d58fd3). On v11, simply disallow builds against mapped catalog
relations by deeming them parallel unsafe.
Author: Peter Geoghegan
Reported-By: "death lock"
Reviewed-By: Tom Lane, Amit Kapila
Bug: #15309
Discussion: https://postgr.es/m/153329671686.1405.18298309097348420351@wrigleys.postgresql.org
Backpatch: 11-, where parallel CREATE INDEX was introduced.
Historically, we've allowed auxiliary processes to take buffer pins without
tracking them in a ResourceOwner. However, that creates problems for error
recovery. In particular, we've seen multiple reports of assertion crashes
in the startup process when it gets an error while holding a buffer pin,
as for example if it gets ENOSPC during a write. In a non-assert build,
the process would simply exit without releasing the pin at all. We've
gotten away with that so far just because a failure exit of the startup
process translates to a database crash anyhow; but any similar behavior
in other aux processes could result in stuck pins and subsequent problems
in vacuum.
To improve this, institute a policy that we must *always* have a resowner
backing any attempt to pin a buffer, which we can enforce just by removing
the previous special-case code in resowner.c. Add infrastructure to make
it easy to create a process-lifespan AuxProcessResourceOwner and clear
out its contents at appropriate times. Replace existing ad-hoc resowner
management in bgwriter.c and other aux processes with that. (Thus, while
the startup process gains a resowner where it had none at all before, some
other aux process types are replacing an ad-hoc resowner with this code.)
Also use the AuxProcessResourceOwner to manage buffer pins taken during
StartupXLOG and ShutdownXLOG, even when those are being run in a bootstrap
process or a standalone backend rather than a true auxiliary process.
In passing, remove some other ad-hoc resource owner creations that had
gotten cargo-culted into various other places. As far as I can tell
that was all unnecessary, and if it had been necessary it was incomplete,
due to lacking any provision for clearing those resowners later.
(Also worth noting in this connection is that a process that hasn't called
InitBufferPoolBackend has no business accessing buffers; so there's more
to do than just add the resowner if we want to touch buffers in processes
not covered by this patch.)
Although this fixes a very old bug, no back-patch, because there's no
evidence of any significant problem in non-assert builds.
Patch by me, pursuant to a report from Justin Pryzby. Thanks to
Robert Haas and Kyotaro Horiguchi for reviews.
Discussion: https://postgr.es/m/20180627233939.GA10276@telsasoft.com
PostgreSQL nowadays offers some kind of dynamic shared memory feature on
all supported platforms. Having the choice of "none" prevents us from
relying on DSM in core features. So this patch removes the choice of
"none".
Author: Kyotaro Horiguchi <horiguchi.kyotaro@lab.ntt.co.jp>
THis adds a "flags" field to the BackgroundWorkerInitializeConnection()
and BackgroundWorkerInitializeConnectionByOid(). For now only one flag,
BGWORKER_BYPASS_ALLOWCONN, is defined, which allows the worker to ignore
datallowconn.
Commit 445dbd82a basically missed the point of commit d46633506,
which was that we shouldn't allow shm_toc_lookup() failure to lead
to a core dump or assertion crash, because the odds of such a
failure should never be considered negligible. It's correct that
we can't expect the PARALLEL_KEY_ERROR_QUEUE TOC entry to be there
if we have no workers. But if we have no workers, we're not going
to do anything in this function with the lookup result anyway,
so let's just skip it. That lets the code use the easy-to-prove-safe
noError=false case, rather than anything requiring effort to review.
Back-patch to v10, like the previous commit.
Discussion: https://postgr.es/m/3647.1517601675@sss.pgh.pa.us
To make this work, tuplesort.c and logtape.c must also support
parallelism, so this patch adds that infrastructure and then applies
it to the particular case of parallel btree index builds. Testing
to date shows that this can often be 2-3x faster than a serial
index build.
The model for deciding how many workers to use is fairly primitive
at present, but it's better than not having the feature. We can
refine it as we get more experience.
Peter Geoghegan with some help from Rushabh Lathia. While Heikki
Linnakangas is not an author of this patch, he wrote other patches
without which this feature would not have been possible, and
therefore the release notes should possibly credit him as an author
of this feature. Reviewed by Claudio Freire, Heikki Linnakangas,
Thomas Munro, Tels, Amit Kapila, me.
Discussion: http://postgr.es/m/CAM3SWZQKM=Pzc=CAHzRixKjp2eO5Q0Jg1SoFQqeXFQ647JiwqQ@mail.gmail.com
Discussion: http://postgr.es/m/CAH2-Wz=AxWqDoVvGU7dq856S4r6sJAj6DBn7VMtigkB33N5eyg@mail.gmail.com
Once this function has been called, we know that all workers have
started and attached to their error queues -- so if any of them
subsequently exit uncleanly, we'll be sure to throw an ERROR promptly.
Otherwise, users of the ParallelContext machinery must be careful not
to wait forever for a worker that has failed to start. Parallel query
manages to work without needing this for reasons explained in new
comments added by this patch, but it's a useful primitive for other
parallel operations, such as the pending patch to make creating a
btree index run in parallel.
Amit Kapila, revised by me. Additional review by Peter Geoghegan.
Discussion: http://postgr.es/m/CAA4eK1+e2MzyouF5bg=OtyhDSX+=Ao=3htN=T-r_6s3gCtKFiw@mail.gmail.com
Commit 28724fd90d fixed things so that
if a background worker fails to start due to fork() failure or because
it is terminated before startup succeeds, BGWH_STOPPED will be
reported. However, that only helps if the code that uses the
background worker machinery notices the change in status, and the code
in parallel.c did not.
To fix that, do two things. First, make sure that when a worker
exits, it triggers the leader to read from error queues. That way, if
a worker which has attached to an error queue exits uncleanly, the
leader is sure to throw some error, either the contents of the
ErrorResponse sent by the worker, or "lost connection to parallel
worker" if it exited without sending one. To cover the case where
the worker never starts up in the first place or exits before
attaching to the error queue, the ParallelContext now keeps track
of which workers have sent at least one message via the error
queue. A worker which sends no messages by the time the parallel
operation finishes will be checked to see whether it exited before
attaching to the error queue; if so, a new error message, "parallel
worker failed to initialize", will be reported. If not, we'll
continue to wait until it either starts up and exits cleanly, starts
up and exits uncleanly, or fails to start, and then take the
appropriate action.
Patch by me, reviewed by Amit Kapila.
Discussion: http://postgr.es/m/CA+TgmoYnBgXgdTu6wk5YPdWhmgabYc9nY_pFLq=tB=FSLYkD8Q@mail.gmail.com
Commit d466335064 changed things so
that shm_toc_lookup would fail with an error rather than silently
returning NULL in the hope that such failures would be reported
in a useful way rather than via a system crash. However, it
overlooked the fact that the lookup of PARALLEL_KEY_ERROR_QUEUE
in ReinitializeParallelDSM is expected to fail when no DSM segment
was created in the first place; in that case, we end up with a
backend-private memory segment that still contains an entry for
PARALLEL_KEY_FIXED but no others. Consequently a benign failure
to initialize parallelism can escalate into an elog(ERROR);
repair.
Discussion: http://postgr.es/m/CA+Tgmob8LFw55DzH1QEREpBEA9RJ_W_amhBFCVZ6WMwUhVpOqg@mail.gmail.com
Without this fix, dropping a role can sometimes result in parallel
query failures in sessions that have used "SET ROLE" to assume the
dropped role, even if that setting isn't active any more.
Report by Pavan Deolasee. Patch by Amit Kapila, reviewed by me.
Discussion: http://postgr.es/m/CABOikdOomRcZsLsLK+Z+qENM1zxyaWnAvFh3MJZzZnnKiF+REg@mail.gmail.com
Add bgw_type field to background worker structure. It is intended to be
set to the same value for all workers of the same type, so they can be
grouped in pg_stat_activity, for example.
The backend_type column in pg_stat_activity now shows bgw_type for a
background worker. The ps listing also no longer calls out that a
process is a background worker but just show the bgw_type. That way,
being a background worker is more of an implementation detail now that
is not shown to the user. However, most log messages still refer to
'background worker "%s"'; otherwise constructing sensible and
translatable log messages would become tricky.
Reviewed-by: Michael Paquier <michael.paquier@gmail.com>
Reviewed-by: Daniel Gustafsson <daniel@yesql.se>
Tuples can have type RECORDOID and a typmod number that identifies a blessed
TupleDesc in a backend-private cache. To support the sharing of such tuples
through shared memory and temporary files, provide a typmod registry in
shared memory.
To achieve that, introduce per-session DSM segments, created on demand when a
backend first runs a parallel query. The per-session DSM segment has a
table-of-contents just like the per-query DSM segment, and initially the
contents are a shared record typmod registry and a DSA area to provide the
space it needs to grow.
State relating to the current session is accessed via a Session object
reached through global variable CurrentSession that may require significant
redesign further down the road as we figure out what else needs to be shared
or remodelled.
Author: Thomas Munro
Reviewed-By: Andres Freund
Discussion: https://postgr.es/m/CAEepm=0ZtQ-SpsgCyzzYpsXS6e=kZWqk3g5Ygn3MDV7A8dabUA@mail.gmail.com
The logic around shm_mq_detach was a few bricks shy of a load, because
(contrary to the comments for shm_mq_attach) all it did was update the
shared shm_mq state. That left us leaking a bit of process-local
memory, but much worse, the on_dsm_detach callback for shm_mq_detach
was still armed. That means that whenever we ultimately detach from
the DSM segment, we'd run shm_mq_detach again for already-detached,
possibly long-dead queues. This accidentally fails to fail today,
because we only ever re-use a shm_mq's memory for another shm_mq, and
multiple detach attempts on the last such shm_mq are fairly harmless.
But it's gonna bite us someday, so let's clean it up.
To do that, change shm_mq_detach's API so it takes a shm_mq_handle
not the underlying shm_mq. This makes the callers simpler in most
cases anyway. Also fix a few places in parallel.c that were just
pfree'ing the handle structs rather than doing proper cleanup.
Back-patch to v10 because of the risk that the revenant shm_mq_detach
callbacks would cause a live bug sometime. Since this is an API
change, it's too late to do it in 9.6. (We could make a variant
patch that preserves API, but I'm not excited enough to do that.)
Discussion: https://postgr.es/m/8670.1504192177@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
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
The larger part of this patch replaces usages of MyProc->procLatch
with MyLatch. The latter works even early during backend startup,
where MyProc->procLatch doesn't yet. While the affected code
shouldn't run in cases where it's not initialized, it might get copied
into places where it might. Using MyLatch is simpler and a bit faster
to boot, so there's little point to stick with the previous coding.
While doing so I noticed some weaknesses around newly introduced uses
of latches that could lead to missed events, and an omitted
CHECK_FOR_INTERRUPTS() call in worker_spi.
As all the actual bugs are in v10 code, there doesn't seem to be
sufficient reason to backpatch this.
Author: Andres Freund
Discussion:
https://postgr.es/m/20170606195321.sjmenrfgl2nu6j63@alap3.anarazel.dehttps://postgr.es/m/20170606210405.sim3yl6vpudhmufo@alap3.anarazel.de
Backpatch: -
Given the possibility of race conditions and so on, it seems entirely
unsafe to just assume that shm_toc_lookup() always finds the key it's
looking for --- but that was exactly what all but one call site were
doing. To fix, add a "bool noError" argument, similarly to what we
have in many other functions, and throw an error on an unexpected
lookup failure. Remove now-redundant Asserts that a rather random
subset of call sites had.
I doubt this will throw any light on buildfarm member lorikeet's
recent failures, because if an unnoticed lookup failure were involved,
you'd kind of expect a null-pointer-dereference crash rather than the
observed symptom. But you never know ... and this is better coding
practice even if it never catches anything.
Discussion: https://postgr.es/m/9697.1496675981@sss.pgh.pa.us
Coverity complained because bgw.bgw_extra wasn't being filled in by
ApplyLauncherRegister(). The most future-proof fix is to memset the
whole BackgroundWorker struct to zeroes. While at it, let's apply the
same coding rule to other places that set up BackgroundWorker structs;
four out of five had the same or related issues.
We'd already recognized that we can't pass function pointers across process
boundaries for functions in loadable modules, since a shared library could
get loaded at different addresses in different processes. But actually the
practice doesn't work for functions in the core backend either, if we're
using EXEC_BACKEND. This is the cause of recent failures on buildfarm
member culicidae. Switch to passing a string function name in all cases.
Something like this needs to be back-patched into 9.6, but let's see
if the buildfarm likes it first.
Petr Jelinek, with a bunch of basically-cosmetic adjustments by me
Discussion: https://postgr.es/m/548f9c1d-eafa-e3fa-9da8-f0cc2f654e60@2ndquadrant.com
On EXEC_BACKEND builds, this can fail if ASLR is in use.
Backpatch to 9.5. On master, completely remove the bgw_main field
completely, since there is no situation in which it is safe for an
EXEC_BACKEND build. On 9.6 and 9.5, leave the field intact to avoid
breaking things for third-party code that doesn't care about working
under EXEC_BACKEND. Prior to 9.5, there are no in-core bgworker
entrypoints.
Petr Jelinek, reviewed by me.
Discussion: http://postgr.es/m/09d8ad33-4287-a09b-a77f-77f8761adb5e@2ndquadrant.com
Increase the default value of the existing max_worker_processes GUC
from 8 to 16, and add a new max_parallel_workers GUC with a maximum
of 8. This way, even if the maximum amount of parallel query is
happening, there is still room for background workers that do other
things, as originally envisioned when max_worker_processes was added.
Julien Rouhaud, reviewed by Amit Kapila and by revised by me.
WaitLatch, WaitLatchOrSocket, and WaitEventSetWait now taken an
additional wait_event_info parameter; legal values are defined in
pgstat.h. This makes it possible to uniquely identify every point in
the core code where we are waiting for a latch; extensions can pass
WAIT_EXTENSION.
Because latches were the major wait primitive not previously covered
by this patch, it is now possible to see information in
pg_stat_activity on a large number of important wait events not
previously addressed, such as ClientRead, ClientWrite, and SyncRep.
Unfortunately, many of the wait events added by this patch will fail
to appear in pg_stat_activity because they're only used in background
processes which don't currently appear in pg_stat_activity. We should
fix this either by creating a separate view for such information, or
else by deciding to include them in pg_stat_activity after all.
Michael Paquier and Robert Haas, reviewed by Alexander Korotkov and
Thomas Munro.
I found that half a dozen (nearly 5%) of our AllocSetContextCreate calls
had typos in the context-sizing parameters. While none of these led to
especially significant problems, they did create minor inefficiencies,
and it's now clear that expecting people to copy-and-paste those calls
accurately is not a great idea. Let's reduce the risk of future errors
by introducing single macros that encapsulate the common use-cases.
Three such macros are enough to cover all but two special-purpose contexts;
those two calls can be left as-is, I think.
While this patch doesn't in itself improve matters for third-party
extensions, it doesn't break anything for them either, and they can
gradually adopt the simplified notation over time.
In passing, change TopMemoryContext to use the default allocation
parameters. Formerly it could only be extended 8K at a time. That was
probably reasonable when this code was written; but nowadays we create
many more contexts than we did then, so that it's not unusual to have a
couple hundred K in TopMemoryContext, even without considering various
dubious code that sticks other things there. There seems no good reason
not to let it use growing blocks like most other contexts.
Back-patch to 9.6, mostly because that's still close enough to HEAD that
it's easy to do so, and keeping the branches in sync can be expected to
avoid some future back-patching pain. The bugs fixed by these changes
don't seem to be significant enough to justify fixing them further back.
Discussion: <21072.1472321324@sss.pgh.pa.us>
HandleParallelMessages leaked memory into the caller's context. Since it's
called from ProcessInterrupts, there is basically zero certainty as to what
CurrentMemoryContext is, which means we could be leaking into long-lived
contexts. Over the processing of many worker messages that would grow to
be a problem. Things could be even worse than just a leak, if we happened
to service the interrupt while ErrorContext is current: elog.c thinks it
can reset that on its own whim, possibly yanking storage out from under
HandleParallelMessages.
Give HandleParallelMessages its own dedicated context instead, which we can
reset during each call to ensure there's no accumulation of wasted memory.
Discussion: <16610.1472222135@sss.pgh.pa.us>
The previous coding here was capable of adding a "parallel worker" context
line to errors that were not, in fact, returned from a parallel worker.
Instead of using an errcontext callback to add that annotation, just paste
it onto the message by hand; this looks uglier but is more reliable.
Discussion: <19757.1472151987@sss.pgh.pa.us>
As noted by Alvaro, there are CHECK_FOR_INTERRUPTS() calls in the shm_mq.c
functions called by HandleParallelMessages(). I believe they're all
unreachable since we always pass nowait = true, but it doesn't seem like
a great idea to assume that no such call will ever be reachable from
HandleParallelMessages(). If that did happen, there would be a risk of a
recursive call to HandleParallelMessages(), which it does not appear to be
designed for --- for example, there's nothing that would prevent
out-of-order processing of received messages. And certainly such cases
cannot easily be tested. So let's prevent it by holding off interrupts for
the duration of the function. Back-patch to 9.5 which contains identical
code.
Discussion: <14869.1470083848@sss.pgh.pa.us>
ParallelMessagePending *must* be marked volatile, because it's set
by a signal handler. On the other hand, it's pointless for
HandleParallelMessageInterrupt to save/restore errno; that must be,
and is, done at the outer level of the SIGUSR1 signal handler.
Calling CHECK_FOR_INTERRUPTS() inside HandleParallelMessages, which itself
is called from CHECK_FOR_INTERRUPTS(), seems both useless and hazardous.
The comment claiming that this is needed to handle the error queue going
away is certainly misguided, in any case.
Improve a couple of error message texts, and use
ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE to report loss of parallel worker
connection, since that's what's used in e.g. tqueue.c. (Maybe it would be
worth inventing a dedicated ERRCODE for this type of failure? But I do not
think ERRCODE_INTERNAL_ERROR is appropriate.)
Minor stylistic cleanups.
Previously, workers sent data to the leader using the client encoding.
That mostly worked, but the leader the converted the data back to the
server encoding. Since not all encoding conversions are reversible,
that could provoke failures. Fix by using the database encoding for
all communication between worker and leader.
Also, while temporary changes to GUC settings, as from the SET clause
of a function, are in general OK for parallel query, changing
client_encoding this way inside of a parallel worker is not OK.
Previously, that would have confused the leader; with these changes,
it would not confuse the leader, but it wouldn't do anything either.
So refuse such changes in parallel workers.
Also, the previous code naively assumed that when it received a
NotifyResonse from the worker, it could pass that directly back to the
user. But now that worker-to-leader communication always uses the
database encoding, that's clearly no longer correct - though,
actually, the old way was always broken for V2 clients. So
disassemble and reconstitute the message instead.
Issues reported by Peter Eisentraut. Patch by me, reviewed by
Peter Eisentraut.
The fact that no workers were successfully launched in the previous
iteration does not excuse us from setting up properly to try again.
This appears to explain crashes I saw in parallel regression testing
due to error_mqh being NULL when it shouldn't be.
Minor other cosmetic fixes too.
Transmit the leader's temp-namespace state to workers. This is important
because without it, the workers do not really have the same search path
as the leader. For example, there is no good reason (and no extant code
either) to prevent a worker from executing a temp function that the
leader created previously; but as things stood it would fail to find the
temp function, and then either fail or execute the wrong function entirely.
We still prohibit a worker from creating a temp namespace on its own.
In effect, a worker can only see the session's temp namespace if the leader
had created it before starting the worker, which seems like the right
semantics.
Also, transmit the leader's BackendId to workers, and arrange for workers
to use that when determining the physical file path of a temp relation
belonging to their session. While the original intent was to prevent such
accesses entirely, there were a number of holes in that, notably in places
like dbsize.c which assume they can safely access temp rels of other
sessions anyway. We might as well get this right, as a small down payment
on someday allowing workers to access the leader's temp tables. (With
this change, directly using "MyBackendId" as a relation or buffer backend
ID is deprecated; you should use BackendIdForTempRelations() instead.
I left a couple of such uses alone though, as they're not going to be
reachable in parallel workers until we do something about localbuf.c.)
Move the thou-shalt-not-access-thy-leader's-temp-tables prohibition down
into localbuf.c, which is where it actually matters, instead of having it
in relation_open(). This amounts to recognizing that access to temp
tables' catalog entries is perfectly safe in a worker, it's only the data
in local buffers that is problematic.
Having done all that, we can get rid of the test in has_parallel_hazard()
that says that use of a temp table's rowtype is unsafe in parallel workers.
That test was unduly expensive, and if we really did need such a
prohibition, that was not even close to being a bulletproof guard for it.
(For example, any user-defined function executed in a parallel worker
might have attempted such access.)
That way, if the result overflows size_t, you'll get an error instead
of undefined behavior, which seems like a plus. This also has the
effect of casting the number of workers from int to Size, which is
better because it's harder to overflow int than size_t.
Dilip Kumar reported this issue and provided a patch upon which this
patch is based, but his version did use mul_size.
Originally, we didn't have nworkers_launched, so code that used parallel
contexts had to be preprared for the possibility that not all of the
workers requested actually got launched. But now we can count on knowing
the number of workers that were successfully launched, which can shave
off a few cycles and simplify some code slightly.
Amit Kapila, reviewed by Haribabu Kommi, per a suggestion from Peter
Geoghegan.
When force_parallel_mode = true, we enable the parallel mode restrictions
for all queries for which this is believed to be safe. For the subset of
those queries believed to be safe to run entirely within a worker, we spin
up a worker and run the query there instead of running it in the
original process. When force_parallel_mode = regress, make additional
changes to allow the regression tests to run cleanly even though parallel
workers have been injected under the hood.
Taken together, this facilitates both better user testing and better
regression testing of the parallelism code.
Robert Haas, with help from Amit Kapila and Rushabh Lathia.
For locking purposes, we now regard heavyweight locks as mutually
non-conflicting between cooperating parallel processes. There are some
possible pitfalls to this approach that are not to be taken lightly,
but it works OK for now and can be changed later if we find a better
approach. Without this, it's very easy for parallel queries to
silently self-deadlock if the user backend holds strong relation locks.
Robert Haas, with help from Amit Kapila. Thanks to Noah Misch and
Andres Freund for extensive discussion of possible issues with this
approach.
Up until now, the total amount of data that could be passed to a
background worker at startup was one datum, which can be a small as
4 bytes on some systems. That's enough to pass a dsm_handle or an
array index, but not much else. Add a bgw_extra flag to the
BackgroundWorker struct, allowing up to 128 bytes to be passed to
a new worker on any platform.
Use this to fix a problem I recently discovered with the parallel
context machinery added in 9.5: the master assigns each worker an
array index, and each worker subsequently assigns itself an array
index, and there's nothing to guarantee that the two sets of indexes
match, leading to chaos.
Normally, I would not back-patch the change to add bgw_extra, since it
is basically a feature addition. However, since 9.5 is still in beta
and there seems to be no other sensible way to repair the broken
parallel context machinery, back-patch to 9.5. Existing background
worker code can ignore the bgw_extra field without a problem, but
might need to be recompiled since the structure size has changed.
Report and patch by me. Review by Amit Kapila.
Commit b0b0d84b3d purported to make it
possible to relaunch workers using the same parallel context, but it had
an unpleasant race condition: we might reinitialize after the workers
have sent their last control message but before they have dettached the
DSM, leaving to crashes. Repair by introducing a new ParallelContext
operation, ReinitializeParallelDSM.
Adjust execParallel.c to use this new support, so that we can rescan a
Gather node by relaunching workers but without needing to recreate the
DSM.
Amit Kapila, with some adjustments by me. Extracted from latest parallel
sequential scan patch.
This may allow some callers to avoid the overhead involved in tearing
down a parallel context and then setting up a new one, which means
releasing the DSM and then allocating and populating a new one. I
suspect we'll want to revise the Gather node to make use of this new
capability, but even if not it may be useful elsewhere and requires
very little additional code.
In order for this to be safe, the code which hands true serializability
will need to taught that the SIRead locks taken by a parallel worker
pertain to the same transaction as those taken by the parallel leader.
Some further changes may be needed as well. Until the necessary
adaptations are made, don't generate parallel plans in serializable
mode, and if a previously-generated parallel plan is used after
serializable mode has been activated, run it serially.
This fixes a bug in commit 7aea8e4f2d.
check_role() tries to verify that the user has permission to become the
requested role, but this is inappropriate in a parallel worker, which
needs to exactly recreate the master's authorization settings. So skip
the check in that case.
This fixes a bug in commit 924bcf4f16.
Starting a parallel worker transaction changes our notion of which XIDs
are in-progress or committed, and our notion of the current command
counter ID. Therefore, our view of these caches prior to starting
this transaction may no longer valid. Defend against that by clearing
them.
This fixes a bug in commit 924bcf4f16.
Thomas Munro
Peter Geoghegan
Tom Lane
Peter Eisentraut
Magnus Hagander
Robert Haas
Bruce Momjian
Andres Freund