Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commit4cff100

Browse files
committed
Fix parallel pg_restore to handle comments on POST_DATA items correctly.
The previous coding would try to process all SECTION_NONE items in theinitial sequential-restore pass, which failed if they were dependencies ofnot-yet-restored items. Fix by postponing such items into the parallelprocessing pass once we have skipped any non-PRE_DATA item.Back-patch into 9.0; the original parallel-restore coding in 8.4 did nothave this bug, so no need to change it.Report and diagnosis by Arnd Hannemann.
1 parent472f608 commit4cff100

File tree

1 file changed

+50
-15
lines changed

1 file changed

+50
-15
lines changed

‎src/bin/pg_dump/pg_backup_archiver.c

Lines changed: 50 additions & 15 deletions
Original file line numberDiff line numberDiff line change
@@ -3218,12 +3218,12 @@ dumpTimestamp(ArchiveHandle *AH, const char *msg, time_t tim)
32183218
* Main engine for parallel restore.
32193219
*
32203220
* Work is done in three phases.
3221-
* First we processtocEntries until we come to one that is marked
3222-
*SECTION_DATA or SECTION_POST_DATA, in a single connection, just as for a
3223-
*standard restore. Second we process the remaining non-ACL steps in
3224-
*parallel worker children (threads on Windows, processes on Unix), each of
3225-
*which connects separately to the database. Finally we process all the ACL
3226-
*entries in a single connection (that happens back inRestoreArchive).
3221+
* First we processall SECTION_PRE_DATA tocEntries, in a single connection,
3222+
*just as for a standard restore. Second we process the remaining non-ACL
3223+
*steps in parallel worker children (threads on Windows, processes on Unix),
3224+
*each of which connects separately to the database. Finally we process all
3225+
*the ACL entries in a single connection (that happens back in
3226+
* RestoreArchive).
32273227
*/
32283228
staticvoid
32293229
restore_toc_entries_parallel(ArchiveHandle*AH)
@@ -3233,6 +3233,7 @@ restore_toc_entries_parallel(ArchiveHandle *AH)
32333233
ParallelSlot*slots;
32343234
intwork_status;
32353235
intnext_slot;
3236+
boolskipped_some;
32363237
TocEntrypending_list;
32373238
TocEntryready_list;
32383239
TocEntry*next_work_item;
@@ -3262,12 +3263,31 @@ restore_toc_entries_parallel(ArchiveHandle *AH)
32623263
* showing all the dependencies of SECTION_PRE_DATA items, so we do not
32633264
* risk trying to process them out-of-order.
32643265
*/
3266+
skipped_some= false;
32653267
for (next_work_item=AH->toc->next;next_work_item!=AH->toc;next_work_item=next_work_item->next)
32663268
{
3267-
/* Non-PRE_DATA items are just ignored for now */
3268-
if (next_work_item->section==SECTION_DATA||
3269-
next_work_item->section==SECTION_POST_DATA)
3270-
continue;
3269+
/* NB: process-or-continue logic must be the inverse of loop below */
3270+
if (next_work_item->section!=SECTION_PRE_DATA)
3271+
{
3272+
/* DATA and POST_DATA items are just ignored for now */
3273+
if (next_work_item->section==SECTION_DATA||
3274+
next_work_item->section==SECTION_POST_DATA)
3275+
{
3276+
skipped_some= true;
3277+
continue;
3278+
}
3279+
else
3280+
{
3281+
/*
3282+
* SECTION_NONE items, such as comments, can be processed now
3283+
* if we are still in the PRE_DATA part of the archive. Once
3284+
* we've skipped any items, we have to consider whether the
3285+
* comment's dependencies are satisfied, so skip it for now.
3286+
*/
3287+
if (skipped_some)
3288+
continue;
3289+
}
3290+
}
32713291

32723292
ahlog(AH,1,"processing item %d %s %s\n",
32733293
next_work_item->dumpId,
@@ -3310,17 +3330,32 @@ restore_toc_entries_parallel(ArchiveHandle *AH)
33103330
*/
33113331
par_list_header_init(&pending_list);
33123332
par_list_header_init(&ready_list);
3333+
skipped_some= false;
33133334
for (next_work_item=AH->toc->next;next_work_item!=AH->toc;next_work_item=next_work_item->next)
33143335
{
3315-
/* All PRE_DATA items were dealt with above */
3336+
/* NB: process-or-continue logic must be the inverse of loop above */
3337+
if (next_work_item->section==SECTION_PRE_DATA)
3338+
{
3339+
/* All PRE_DATA items were dealt with above */
3340+
continue;
3341+
}
33163342
if (next_work_item->section==SECTION_DATA||
33173343
next_work_item->section==SECTION_POST_DATA)
33183344
{
3319-
if (next_work_item->depCount>0)
3320-
par_list_append(&pending_list,next_work_item);
3321-
else
3322-
par_list_append(&ready_list,next_work_item);
3345+
/* set this flag at same point that previous loop did */
3346+
skipped_some= true;
33233347
}
3348+
else
3349+
{
3350+
/* SECTION_NONE items must be processed if previous loop didn't */
3351+
if (!skipped_some)
3352+
continue;
3353+
}
3354+
3355+
if (next_work_item->depCount>0)
3356+
par_list_append(&pending_list,next_work_item);
3357+
else
3358+
par_list_append(&ready_list,next_work_item);
33243359
}
33253360

33263361
/*

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp