pg.ddx.io  pgsql-hackers@postgresql.org mailing list archive  
help / color / mirror / Atom feed
From: Álvaro Herrera <alvherre@alvh.no-ip.org>
To: Masahiko Sawada <sawada.mshk@gmail.com>
Cc: Hamid Akhtar <hamid.akhtar@gmail.com>
Cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: {CREATE INDEX, REINDEX} CONCURRENTLY improvements
Date: Fri, 15 Jan 2021 11:29:26 -0300
Message-ID: <20210115142926.GA19300@alvherre.pgsql> (raw)
In-Reply-To: <20210115133858.GA18931@alvherre.pgsql>

So one last remaining improvement was to have VACUUM ignore processes
doing CIC and RC to compute the Xid horizon of tuples to remove.  I
think we can do something simple like the attached patch.

-- 
Álvaro Herrera       Valdivia, Chile
"Aprender sin pensar es inútil; pensar sin aprender, peligroso" (Confucio)

Attachments:

vacuum-cic.patch (text/x-diff, 4.9K) download diff --git a/src/backend/storage/ipc/procarray.c b/src/backend/storage/ipc/procarray.c index cf12eda504..f584230b79 100644 --- a/src/backend/storage/ipc/procarray.c +++ b/src/backend/storage/ipc/procarray.c @@ -1601,7 +1601,7 @@ TransactionIdIsActive(TransactionId xid) * well as "internally" by GlobalVisUpdate() (see comment above struct * GlobalVisState). * - * See the definition of ComputedXidHorizonsResult for the various computed + * See the definition of ComputeXidHorizonsResult for the various computed * horizons. * * For VACUUM separate horizons (used to decide which deleted tuples must @@ -1610,7 +1610,12 @@ TransactionIdIsActive(TransactionId xid) * relations that's not required, since only backends in my own database could * ever see the tuples in them. Also, we can ignore concurrently running lazy * VACUUMs because (a) they must be working on other tables, and (b) they - * don't need to do snapshot-based lookups. + * don't need to do snapshot-based lookups. Also, for the non-catalog + * horizon, we can ignore CREATE INDEX CONCURRENTLY and REINDEX CONCURRENTLY + * when they are working on non-partial, non-expressional indexes, for the + * same reasons and because they can't run in transaction blocks. (They are + * not possible to ignore for catalogs, because CIC and RC do some catalog + * operations.) * * This also computes a horizon used to truncate pg_subtrans. For that * backends in all databases have to be considered, and concurrently running @@ -1660,9 +1665,6 @@ ComputeXidHorizons(ComputeXidHorizonsResult *h) bool in_recovery = RecoveryInProgress(); TransactionId *other_xids = ProcGlobal->xids; - /* inferred after ProcArrayLock is released */ - h->catalog_oldest_nonremovable = InvalidTransactionId; - LWLockAcquire(ProcArrayLock, LW_SHARED); h->latest_completed = ShmemVariableCache->latestCompletedXid; @@ -1682,6 +1684,7 @@ ComputeXidHorizons(ComputeXidHorizonsResult *h) h->oldest_considered_running = initial; h->shared_oldest_nonremovable = initial; + h->catalog_oldest_nonremovable = initial; h->data_oldest_nonremovable = initial; /* @@ -1752,7 +1755,7 @@ ComputeXidHorizons(ComputeXidHorizonsResult *h) if (statusFlags & (PROC_IN_VACUUM | PROC_IN_LOGICAL_DECODING)) continue; - /* shared tables need to take backends in all database into account */ + /* shared tables need to take backends in all databases into account */ h->shared_oldest_nonremovable = TransactionIdOlder(h->shared_oldest_nonremovable, xmin); @@ -1768,16 +1771,29 @@ ComputeXidHorizons(ComputeXidHorizonsResult *h) * to prune still needed data away). If the current backend never * connects to a database that is harmless, because * data_oldest_nonremovable will never be utilized. + * + * Additionally, processes doing CREATE INDEX CONCURRENTLY and REINDEX + * CONCURRENTLY on "safe" indexes can be ignored for non-catalog + * horizon. (But not for catalogs: some transactions in CIC/RC do + * catalog updates.) */ if (in_recovery || MyDatabaseId == InvalidOid || proc->databaseId == MyDatabaseId || proc->databaseId == 0) /* always include WalSender */ { - h->data_oldest_nonremovable = - TransactionIdOlder(h->data_oldest_nonremovable, xmin); + if (statusFlags & PROC_IN_SAFE_IC) + h->catalog_oldest_nonremovable = + TransactionIdOlder(h->catalog_oldest_nonremovable, xmin); + else + h->data_oldest_nonremovable = h->catalog_oldest_nonremovable = + TransactionIdOlder(h->data_oldest_nonremovable, xmin); } } + /* catalog horizon should never be later than data */ + Assert(TransactionIdPrecedesOrEquals(h->catalog_oldest_nonremovable, + h->data_oldest_nonremovable)); + /* * If in recovery fetch oldest xid in KnownAssignedXids, will be applied * after lock is released. @@ -1799,6 +1815,8 @@ ComputeXidHorizons(ComputeXidHorizonsResult *h) TransactionIdOlder(h->shared_oldest_nonremovable, kaxmin); h->data_oldest_nonremovable = TransactionIdOlder(h->data_oldest_nonremovable, kaxmin); + h->catalog_oldest_nonremovable = + TransactionIdOlder(h->catalog_oldest_nonremovable, kaxmin); /* temp relations cannot be accessed in recovery */ } else @@ -1825,6 +1843,9 @@ ComputeXidHorizons(ComputeXidHorizonsResult *h) h->data_oldest_nonremovable = TransactionIdRetreatedBy(h->data_oldest_nonremovable, vacuum_defer_cleanup_age); + h->catalog_oldest_nonremovable = + TransactionIdRetreatedBy(h->catalog_oldest_nonremovable, + vacuum_defer_cleanup_age); /* defer doesn't apply to temp relations */ } @@ -1847,7 +1868,6 @@ ComputeXidHorizons(ComputeXidHorizonsResult *h) h->shared_oldest_nonremovable = TransactionIdOlder(h->shared_oldest_nonremovable, h->slot_catalog_xmin); - h->catalog_oldest_nonremovable = h->data_oldest_nonremovable; h->catalog_oldest_nonremovable = TransactionIdOlder(h->catalog_oldest_nonremovable, h->slot_catalog_xmin);

view thread (19+ messages)  latest in thread

Message-ID: <20210115142926.GA19300@alvherre.pgsql>
Permalink:  ../20210115142926.GA19300@alvherre.pgsql/
Also on:    postgresql.org/message-id/20210115142926.GA19300@alvherre.pgsql

 ·  · 

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: pgsql-hackers@postgresql.org
  Cc: alvherre@alvh.no-ip.org, sawada.mshk@gmail.com, hamid.akhtar@gmail.com, pgsql-hackers@lists.postgresql.org
  Subject: Re: {CREATE INDEX, REINDEX} CONCURRENTLY improvements
  In-Reply-To: <20210115142926.GA19300@alvherre.pgsql>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox