This requires adjusting the API for syscache callback functions: they now
get a hash value, not a TID, to identify the target tuple. Most of them
weren't paying any attention to that argument anyway, but plancache did
require a small amount of fixing.
Also, improve performance a trifle by avoiding sending duplicate inval
messages when a heap_update isn't changing the catcache lookup columns.
The latch infrastructure is now capable of detecting all cases where the
walsender loop needs to wake up, so there is no reason to have an arbitrary
timeout.
Also, modify the walsender loop logic to follow the standard pattern of
ResetLatch, test for work to do, WaitLatch. The previous coding was both
hard to follow and buggy: it would sometimes busy-loop despite having
nothing available to do, eg between receipt of a signal and the next time
it was caught up with new WAL, and it also had interesting choices like
deciding to update to WALSNDSTATE_STREAMING on the strength of information
known to be obsolete.
There may be some other places where we should use errdetail_internal,
but they'll have to be evaluated case-by-case. This commit just hits
a bunch of places where invoking gettext is obviously a waste of cycles.
We were using GetConfigOption to collect the old value of each setting,
overlooking the possibility that it didn't exist yet. This does happen
in the case of adding a new entry within a custom variable class, as
exhibited in bug #6097 from Maxim Boguk.
To fix, add a missing_ok parameter to GetConfigOption, but only in 9.1
and HEAD --- it seems possible that some third-party code is using that
function, so changing its API in a minor release would cause problems.
In 9.0, create a near-duplicate function instead.
Per discussion, this structure seems more understandable than what was
there before. Make config.sgml and postgresql.conf.sample agree.
In passing do a bit of editorial work on the variable descriptions.
postmaster.log", or nohup.
There was a small issue with LINUX_OOM_ADJ and silent_mode, namely that with
silent_mode the postmaster process incorrectly used the OOM settings meant
for backend processes. We certainly could've fixed that directly, but since
silent_mode was redundant anyway, we might as well just remove it.
This allows deadlock_timeout to be reduced for transactions that are
particularly likely to be involved in a deadlock, thus detecting it
more quickly. It is also potentially useful as a poor-man's deadlock
priority mechanism: a transaction with a high deadlock_timeout is less
likely to be chosen as the victim than one with a low
deadlock_timeout. Since that could be used to game the system, we
make this PGC_SUSET rather than PGC_USERSET.
At some point, it might be worth thinking about a more explicit
priority mechanism, since using this is far from fool-proof. But
let's see whether there's enough use case to justify the additional
work before we go down that route.
Noah Misch, reviewed by Shigeru Hanada
Failure to distinguish these cases is the real cause behind the recent
reports of Windows builds crashing on 'infinity'::timestamp, which was
directly due to failure to establish a value of timezone_abbreviations
in postmaster child processes. The postmaster had the desired value,
but write_one_nondefault_variable() didn't transmit it to backends.
To fix that, invent a new value PGC_S_DYNAMIC_DEFAULT, and be sure to use
that or PGC_S_ENV_VAR (as appropriate) for "default" settings that are
computed during initialization. (We need both because there's at least
one variable that could receive a value from either source.)
This commit also fixes ProcessConfigFile's failure to restore the correct
default value for certain GUC variables if they are set in postgresql.conf
and then removed/commented out of the file. We have to recompute and
reinstall the value for any GUC variable that could have received a value
from PGC_S_DYNAMIC_DEFAULT or PGC_S_ENV_VAR sources, and there were a
number of oversights. (That whole thing is a crock that needs to be
redesigned, but not today.)
However, I intentionally didn't make it work "exactly right" for the cases
of timezone and log_timezone. The exactly right behavior would involve
running select_default_timezone, which we'd have to do independently in
each postgres process, causing the whole database to become entirely
unresponsive for as much as several seconds. That didn't seem like a good
idea, especially since the variable's removal from postgresql.conf might be
just an accidental edit. Instead the behavior is to adopt the previously
active setting as if it were default.
Note that this patch creates an ABI break for extensions that use any of
the PGC_S_XXX constants; they'll need to be recompiled.
The recent cleanup of GUC assign hooks got rid of the kludge of using
"unknown" as a magic value for timezone and log_timezone. But I forgot
to update the documentation to match, as noted by Martin Pitt.
The previous functions of assign hooks are now split between check hooks
and assign hooks, where the former can fail but the latter shouldn't.
Aside from being conceptually clearer, this approach exposes the
"canonicalized" form of the variable value to guc.c without having to do
an actual assignment. And that lets us fix the problem recently noted by
Bernd Helmle that the auto-tune patch for wal_buffers resulted in bogus
log messages about "parameter "wal_buffers" cannot be changed without
restarting the server". There may be some speed advantage too, because
this design lets hook functions avoid re-parsing variable values when
restoring a previous state after a rollback (they can store a pre-parsed
representation of the value instead). This patch also resolves a
longstanding annoyance about custom error messages from variable assign
hooks: they should modify, not appear separately from, guc.c's own message
about "invalid parameter value".
Also avoid hardcoding the current default state by giving it the name
"on" and replace with a meaningful name that reflects its behaviour.
Coding only, no change in behaviour.
This means one less thing to configure when setting up synchronous
replication, and also avoids some ambiguity around what the behavior
should be when the settings of these variables conflict.
Fujii Masao, with additional hacking by me.
than replication_timeout (a new GUC) milliseconds. The TCP timeout is often
too long, you want the master to notice a dead connection much sooner.
People complained about that in 9.0 too, but with synchronous replication
it's even more important to notice dead connections promptly.
Fujii Masao and Heikki Linnakangas
The maximum value of deadlock_timeout, max_standby_archive_delay,
max_standby_streaming_delay, log_min_duration_statement, and
log_autovacuum_min_duration was INT_MAX/1000 milliseconds, which is
about 35min, which is too short for some practical uses. Raise the
maximum value to INT_MAX; the code that uses the parameters already
supports that just fine.
SyncRepRequested() must check not only the value of the
synchronous_replication GUC but also whether max_wal_senders > 0.
Otherwise, we might end up waiting for sync rep even when there's no
possibility of a standby ever managing to connect. There are some
existing cross-checks to prevent this, but they're not quite sufficient:
the user can start the server with max_wal_senders=0,
synchronous_standby_names='', and synchronous_replication=off and then
subsequent make synchronous_standby_names not empty using pg_ctl reload,
and then SET synchronous_standby=on, leading to an indefinite hang.
Along the way, rename the global variable for the synchronous_replication
GUC to match the name of the GUC itself, for clarity.
Report by Fujii Masao, though I didn't use his patch.
If a standby is broadcasting reply messages and we have named
one or more standbys in synchronous_standby_names then allow
users who set synchronous_replication to wait for commit, which
then provides strict data integrity guarantees. Design avoids
sending and receiving transaction state information so minimises
bookkeeping overheads. We synchronize with the highest priority
standby that is connected and ready to synchronize. Other standbys
can be defined to takeover in case of standby failure.
This version has very strict behaviour; more relaxed options
may be added at a later date.
Simon Riggs and Fujii Masao, with reviews by Yeb Havinga, Jaime
Casanova, Heikki Linnakangas and Robert Haas, plus the assistance
of many other design reviewers.
Standby optionally sends back information about oldestXmin of queries
which is then checked and applied to the WALSender's proc->xmin.
GetOldestXmin() is modified slightly to agree with GetSnapshotData(),
so that all backends on primary include WALSender within their snapshots.
Note this does nothing to change the snapshot xmin on either master or
standby. Feedback piggybacks on the standby reply message.
vacuum_defer_cleanup_age is no longer used on standby, though parameter
still exists on primary, since some use cases still exist.
Simon Riggs, review comments from Fujii Masao, Heikki Linnakangas, Robert Haas
the standby has written, flushed, and applied the WAL. At the moment, this
is for informational purposes only, the values are only shown in
pg_stat_replication system view, but in the future they will also be needed
for synchronous replication.
Extracted from Simon riggs' synchronous replication patch by Robert Haas, with
some tweaking by me.
Until now, our Serializable mode has in fact been what's called Snapshot
Isolation, which allows some anomalies that could not occur in any
serialized ordering of the transactions. This patch fixes that using a
method called Serializable Snapshot Isolation, based on research papers by
Michael J. Cahill (see README-SSI for full references). In Serializable
Snapshot Isolation, transactions run like they do in Snapshot Isolation,
but a predicate lock manager observes the reads and writes performed and
aborts transactions if it detects that an anomaly might occur. This method
produces some false positives, ie. it sometimes aborts transactions even
though there is no anomaly.
To track reads we implement predicate locking, see storage/lmgr/predicate.c.
Whenever a tuple is read, a predicate lock is acquired on the tuple. Shared
memory is finite, so when a transaction takes many tuple-level locks on a
page, the locks are promoted to a single page-level lock, and further to a
single relation level lock if necessary. To lock key values with no matching
tuple, a sequential scan always takes a relation-level lock, and an index
scan acquires a page-level lock that covers the search key, whether or not
there are any matching keys at the moment.
A predicate lock doesn't conflict with any regular locks or with another
predicate locks in the normal sense. They're only used by the predicate lock
manager to detect the danger of anomalies. Only serializable transactions
participate in predicate locking, so there should be no extra overhead for
for other transactions.
Predicate locks can't be released at commit, but must be remembered until
all the transactions that overlapped with it have completed. That means that
we need to remember an unbounded amount of predicate locks, so we apply a
lossy but conservative method of tracking locks for committed transactions.
If we run short of shared memory, we overflow to a new "pg_serial" SLRU
pool.
We don't currently allow Serializable transactions in Hot Standby mode.
That would be hard, because even read-only transactions can cause anomalies
that wouldn't otherwise occur.
Serializable isolation mode now means the new fully serializable level.
Repeatable Read gives you the old Snapshot Isolation level that we have
always had.
Kevin Grittner and Dan Ports, reviewed by Jeff Davis, Heikki Linnakangas and
Anssi Kääriäinen
As in commit fb4c5d2798 on 2011-01-21,
this avoids spurious debug messages and allows idempotent changes at
any time. Along the way, make assign_XactIsoLevel allow idempotent
changes even when not within a subtransaction, to be consistent with
the new coding of assign_transaction_read_only and because there's
no compelling reason to do otherwise.
Kevin Grittner, with some adjustments.
If wal_buffers is initially set to -1 (which is now the default), it's
replaced by 1/32nd of shared_buffers, with a minimum of 8 (the old default)
and a maximum of the XLOG segment size. The allowed range for manual
settings is still from 4 up to whatever will fit in shared memory.
Greg Smith, with implementation correction by me.
The new coding avoids a spurious debug message when a transaction
that has changed the isolation level has been rolled back. It also
allows the property to be freely changed to the current value within
a subtransaction.
Kevin Grittner, with one small change by me.
Recent versions of the Linux system header files cause xlogdefs.h to
believe that open_datasync should be the default sync method, whereas
formerly fdatasync was the default on Linux. open_datasync is a bad
choice, first because it doesn't actually outperform fdatasync (in fact
the reverse), and second because we try to use O_DIRECT with it, causing
failures on certain filesystems (e.g., ext4 with data=journal option).
This part of the patch is largely per a proposal from Marti Raudsepp.
More extensive changes are likely to follow in HEAD, but this is as much
change as we want to back-patch.
Also clean up confusing code and incorrect documentation surrounding the
fsync_writethrough option. Those changes shouldn't result in any actual
behavioral change, but I chose to back-patch them anyway to keep the
branches looking similar in this area.
In 9.0 and HEAD, also do some copy-editing on the WAL Reliability
documentation section.
Back-patch to all supported branches, since any of them might get used
on modern Linux versions.
First, avoid scanning the whole ProcArray once we know there
are at least commit_siblings active; second, skip the check
altogether if commit_siblings = 0.
Greg Smith
This eliminates some crufty, special-purpose code and, as a non-trivial
side benefit, allows recovery.conf parameters to be unquoted.
Dimitri Fontaine, with review and cleanup by Alvaro Herrera, Itagaki
Takahiro, and me.
Rather than considering this result as meaning "unknown", report LONG_MAX.
This won't change what superusers can set max_stack_depth to, but it will
cause InitializeGUCOptions() to set the built-in default to 2MB not 100kB.
The latter seems like a fairly unreasonable interpretation of "infinity".
Per my investigation of odd buildfarm results as well as an old complaint
from Heikki.
Since this should persuade all the buildfarm animals to use a reasonable
stack depth setting during "make check", revert previous patch that dumbed
down a recursive regression test to only 5 levels.
Split the old typenameTypeId() into two functions: A new typenameTypeId() that
returns only a type OID, and typenameTypeIdAndMod() that returns type OID and
typmod. This isolates call sites better that actually care about the typmod.
A long time ago, this didn't work nicely, but it seems to work on all recent
versions of OS X. The blank-pad method is less desirable since it results
in lots of extra space in ps' output. Per Alexey Klyukin.
Per gripe from Fujii Masao, though this is not exactly his proposed patch.
Categorize as DEVELOPER_OPTIONS and set context PGC_SIGHUP, as per Fujii,
but set the default to LOG because higher values aren't really sensible
(see the code for trace_recovery()). Fix the documentation to agree with
the code and to try to explain what the variable actually does. Get rid
of no-op calls trace_recovery(LOG), which accomplish nothing except to
demonstrate that this option confuses even its author.
This allows us to reliably remove all leftover temporary relation
files on cluster startup without reference to system catalogs or WAL;
therefore, we no longer include temporary relations in XLOG_XACT_COMMIT
and XLOG_XACT_ABORT WAL records.
Since these changes require including a backend ID in each
SharedInvalSmgrMsg, the size of the SharedInvalidationMessage.id
field has been reduced from two bytes to one, and the maximum number
of connections has been reduced from INT_MAX / 4 to 2^23-1. It would
be possible to remove these restrictions by increasing the size of
SharedInvalidationMessage by 4 bytes, but right now that doesn't seem
like a good trade-off.
Review by Jaime Casanova and Tom Lane.
struct representing a tree entry, rather than being a separately allocated
piece of storage. This API is at least as clean as the old one (if not
more so --- there were some bizarre choices in there) and it permits a
very substantial memory savings, on the order of 2X in ginbulk.c's usage.
Also, fix minor memory leaks in code called by ginEntryInsert, in
particular in ginInsertValue and entryFillRoot, as well as ginEntryInsert
itself. These leaks resulted in the GIN index build context continuing
to bloat even after we'd filled it to maintenance_work_mem and started
to dump data out to the index.
In combination these fixes restore the GIN index build code to honoring
the maintenance_work_mem limit about as well as it did in 8.4. Speed
seems on par with 8.4 too, maybe even a bit faster, for a non-pathological
case in which HEAD was formerly slower.
Back-patch to 9.0 so we don't have a performance regression from 8.4.
We used to be consistent about this, but my recent patch to add a
restart_after_crash GUC failed to follow the existing convention.
Report and patch from Fujii Masao.
