Skip to content

Wöchentlicher PostgreSQL Newsletter - 25. April 2010

Der Originalartikel befindet sich unter:

== Wöchentlicher PostgreSQL Newsletter - 25. April 2010 ==

SEAPUG und PDXPUG werden einen Stand auf der LFNW haben - komm vorbei!

Simon Riggs und Hannu Krosing werden über PostgreSQL auf der
"Pycon Italia 4" (die italienische Python Konferenz) in Florenz,
Italien am 8. und 9. Mai sprechen.

Die Anmeldung für PGCon 2010 ist offen. Die Plätze sind schnell
vergeben, also schnell anmelden, wenn du hinkommen willst.

== PostgreSQL Produkt Neuigkeiten ==

PostgreSQL Maestro 10.4, ein GUI zum Administrieren und für die
Entwicklung unter Windows, ist erschienen.

DBD::Pg 2.17.0, ein Perl Treiber für PostgreSQL, ist erschienen.

pgpool-II 2.3.3, ein Connection Pooler und mehr,
ist erschienen.

piggly 1.2.0, ein PL/pgsql Code Coverage Werkzeug, ist erschienen.

== PostgreSQL 9.0 Feature der Woche ==

Funktionen können jetzt Defaultwerte für die Parameter haben.

== PostgreSQL Jobs im April ==

== PostgreSQL Lokal ==

Der CfP für CHAR(10) in Oxford, England vom 1. bis 3. Juli 2010
ist offen. Proposals sind bis zum 20. April einzureichen.

Bruce Momjian wird über PostgreSQL Replikationslösungen auf dem
NYPUG Treffen am 20. April um 18:30-20:30 im Penn Plaza sprechen.
Details und Anmeldung unter:

Linuxfest Nordwest 2010 findet in Bellingham, Washington, USA am
24. und 25. April statt. Vortragsreihen, Stände und Sponsoring
sind verfügbar.

Andreas (ads) Scherbaum hält einen Kurs "PostgreSQL im
Unternehmenseinsatz" an der VHS Magdeburg, Deutschland, vom 03. bis 07.
Mai 2010. Details unter:

PgCon 2010 findet vom 20.-21. Mai 2010 in Ottawa statt mit Tutorials
vorher am 18. und 19.

Der CfP für OSBridge ist eröffnet! OSBridge findet vom 1. bis 4.
Juni 2010 in Portland, Oregon statt.

Michael Renner wird auf der AMOOCON, welche vom 4. bis 6. Juni 2010
stattfindet, sprechen.

Das Southeast Linuxfest 2010 findet in Spartanburg, SC, USA am
12. und 13. Juni statt. Stände und Sponsoringmöglichkeiten sind
noch verfügbar.

OSCON wird in Portland, Oregon vom 19. bis 23. Juli 2010 stattfinden.

== PostgreSQL in den News ==

Planet PostgreSQL:

Dieser wöchentliche PostgreSQL Newsletter wurde erstellt von David

Sende Neuigkeiten und Ankündigungen bis Sonntag, 15 Uhr Pazifischer
Zeit. Bitte sende englische Beiträge an, deutsche an, italienische an

== Angewandte Patches ==

Magnus Hagander committed:

- Add wrapper function libpqrcv_PQexec() in the walreceiver that uses
  async libpq to send queries, making the waiting for responses
  interruptible on platforms where PQexec() can't normally be
  interrupted by signals, such as win32.  Fujii Masao and Magnus

Tom Lane committed:

- In pgsql/src/backend/access/nbtree/nbtxlog.c, fix uninitialized
  local variables.  Not sure why gcc doesn't complain about these ---
  maybe because they're effectively unused?  MSVC does complain
  though, per buildfarm.

- Move the responsibility for calling StartupXLOG into InitPostgres,
  for those process types that go through InitPostgres; in particular,
  bootstrap and standalone-backend cases.  This ensures that we have
  set up a PGPROC and done some other basic initialization steps
  (corresponding to the if (IsUnderPostmaster) block in
  AuxiliaryProcessMain) before we attempt to run WAL recovery in a
  standalone backend.  As was discovered last September, this is
  necessary for some corner-case code paths during WAL recovery,
  particularly end-of-WAL cleanup.  Moving the bootstrap case here too
  is not necessary for correctness, but it seems like a good idea
  since it reduces the number of distinct code paths.

- In pgsql/src/backend/replication/walreceiver.c, fix code that
  doesn't work on machines with strict alignment requirements: must
  use memcpy here rather than struct assignment.  In passing,
  rearrange some randomly-ordered declarations to be a tad less

- Arrange for client authentication to occur before we select a
  specific database to connect to. This is necessary for the walsender
  code to work properly (it was previously using an untenable
  assumption that template1 would always be available to connect to).
  This also gets rid of a small security shortcoming that was
  introduced in the original patch to eliminate the flat
  authentication files: before, you could find out whether or not the
  requested database existed even if you couldn't pass the
  authentication checks.  The changes needed to support this are
  mainly just to treat pg_authid and pg_auth_members as nailed
  relations, so that we can read them without having to be able to
  locate real pg_class entries for them.  This mechanism was already
  debugged for pg_database, but we hadn't recognized the value of
  applying it to those catalogs too.  Since the current code doesn't
  have support for accessing toast tables before we've brought up all
  of the relcache, remove pg_authid's toast table to ensure that no
  one can store an out-of-line toasted value of rolpassword.  The case
  seems quite unlikely to occur in practice, and was effectively
  unsupported anyway in the old "flatfiles" implementation.  Update to actually implement the same rules as bootstrap.c does
  for not-nullability of catalog columns.  The previous coding was a
  bit cheesy but worked all right for the previous set of bootstrap
  catalogs.  It does not work for pg_authid, where rolvaliduntil needs
  to be nullable.  Initdb forced due to minor catalog changes (mainly
  the toast table removal).

- Move the check for whether walreceiver has authenticated as a
  superuser from walsender.c, where it didn't really belong, to
  postinit.c where it does belong (and is essentially free, too).

- Fix pg_hba.conf matching so that replication connections only match
  records with database = replication.  The previous coding would
  allow them to match ordinary records too, but that seems like a
  recipe for security breaches.  Improve the messages associated with
  no-such-pg_hba.conf entry to report replication connections as such,
  since that's now a critical aspect of whether the connection
  matches.  Make some cursory improvements in the related
  documentation, too.

- Enforce superuser permissions checks during ALTER ROLE/DATABASE SET,
  rather than during define_custom_variable().  This entails rejecting
  an ALTER command if the target variable doesn't have a known
  (non-placeholder) definition, unless the calling user is superuser.
  When the variable is known, we can correctly apply the rule that
  only superusers can issue ALTER for SUSET parameters.  This allows
  define_custom_variable to apply ALTER's values for SUSET parameters
  at module load time, secure in the knowledge that only a superuser
  could have set the ALTER value.  This change fixes a longstanding
  gotcha in the usage of SUSET-level custom parameters; which is a
  good thing to fix now that plpgsql defines such a parameter.

Simon Riggs committed:

- In pgsql/src/backend/storage/ipc/procarray.c, check
  RecoveryInProgress() while holding ProcArrayLock during snapshots.
  This prevents a rare, yet possible race condition at the exact
  moment of transition from recovery to normal running.

- Add new message for explicit rejection by pg_hba.conf.  Implicit
  rejection retains same message as before.

- Fix oversight in collecting values for cleanup_info records.
  vacuum_log_cleanup_info() now generates log records with a valid
  latestRemovedXid set in all cases.  Also be careful not to zero the
  value when we do a round of vacuuming part-way through
  lazy_scan_heap().  Incidentally, this reduces frequency of conflicts
  in Hot Standby.

- Relax locking during GetCurrentVirtualXIDs().  Earlier improvements
  to handling of btree delete records mean that all snapshot conflicts
  on standby now have a valid, useful latestRemovedXid.  Our earlier
  approach using LW_EXCLUSIVE was useful when we didnt always have a
  valid value, though is no longer useful or necessary.  Asserts added
  to code path to prove and ensure this is the case.  This will reduce
  contention and improve performance of larger Hot Standby servers.

- In pgsql/src/backend/commands/vacuumlazy.c, only send cleanup_info
  messages if VACUUM removes any tuples.  There is no other purpose
  for this message type than to report the latestRemovedXid of removed
  tuples, prior to index scans.  Removes overlooked path for sending
  invalid latestRemovedXid.  Fixes buildfarm failure on centaur.

- Further reductions in Hot Standby conflict processing.  These come
  from the realistion that HEAP2_CLEAN records don't always remove
  user visible data, so conflict processing for them can be skipped.
  Confirm validity using Assert checks, clarify circumstances under
  which we log heap_cleanup_info records.  Tuning arises from bug
  fixing of earlier safety check failures.

- Optimise btree delete processing when no active backends.  Clarify
  comments, downgrade a message to DEBUG and remove some debug
  counters.  Direct from ideas by Heikki Linnakangas.

- Make CheckRequiredParameterValues() depend upon correct combination
  of parameters.  Fix bug report by Robert Haas that error message and
  hint was incorrect if wrong mode parameters specified on master.
  Internal changes only.  Proposals for parameter simplification on
  master/primary still under way.

- Previous patch revoked following objections.

- Add missing optimizer hooks for function cost and number of rows.
  Closely follow design of other optimizer hooks: if hook exists
  retrieve value from plugin; if still not set then get from cache.

- Patch revoked because of objections.

Robert Haas committed:

- Update docs as to when WAL logging can be skipped.  In 8.4 and
  prior, WAL-logging could potentially be skipped whenever
  archive_mode=off.  With streaming replication, this is now true only
  if max_wal_senders=0.  Fujii Masao, with light copyediting by me.

- Rename standby_keep_segments to wal_keep_segments.  Also, make the
  name of the GUC and the name of the backing variable match.  Alnong
  the way, clean up a couple of slight typographical errors in the
  related docs.

- Fix various instances of "the the".  Two of these were pointed out
  by Erik Rijkers; the rest I found.

ITAGAKI Takahiro committed:

- In pgsql/src/backend/utils/adt/pg_locale.c, fix encoding issue when
  lc_monetary or lc_numeric are different encoding from lc_ctype, that
  could happen on Windows.  We need to change lc_ctype together with
  lc_monetary or lc_numeric, and convert strings in lconv from
  lc_ctype encoding to the database encoding.  The bug reported by
  Mikko, original patch by Hiroshi Inoue, with changes by Bruce and

Peter Eisentraut committed:

- In pgsql/src/backend/utils/misc/guc.c, fix message punctuation.

Bruce Momjian committed:

- In pgsql/src/backend/utils/adt/pg_locale.c, add C comments for
  recent to_char('L') fix for Win32.

== Abgelehnte Patches (bis jetzt) ==

Simon Riggs's patch to speed up ALTER TABLE...SET TABLESPACE.  Timing
is not right for 9.0.

Nikhil Sontakke's patch to estimate the costs of SRFs.

Simon Riggs's patch to make the "ALL" in GRANT ... ON ALL optional.

== Eingesandte Patches ==

Robert Haas sent in a patch to cause walsender not to consume
superuser_reserved_connection slots.

Simon Riggs sent in a patch to implement prefetching during replay of
btree delete records.  The patch is intended to fix a performance
issue reported by Erik Rijkers where a Hot Standby slave's performance
is much lower than on the master's.

Bruce Momjian sent in a patch per prior discussion to re-order the
timestamps displayed in the pg_stat_activity view.

Robert Haas sent in a WIP patch to push isTemp down into the smgr
layer.  This is intended to be infrastructure for global temporary
tables per the SQL standard.


No Trackbacks


Display comments as Linear | Threaded

No comments

Add Comment

Enclosing asterisks marks text as bold (*word*), underscore are made via _word_.
E-Mail addresses will not be displayed and will only be used for E-Mail notifications.
To leave a comment you must approve it via e-mail, which will be sent to your address after submission.
Form options