\\r\\n> We already allow users to specify the degree of parallelism for all\\r\\n> the parallel operations via guc's max_parallel_maintenance_workers,\\r\\n> max_parallel_workers_per_gather, then we have a reloption\\r\\n> parallel_workers and vacuum command has the parallel option where\\r\\n> users can specify the number of workers that can be used for\\r\\n> parallelism. The parallelism considers these as hints but decides\\r\\n> parallelism based on some other parameters like if there are that many\\r\\n> workers available, etc. Why the users would expect differently for\\r\\n> parallel DML?\\r\\n\\r\\nI agree that the user would want to specify the degree of parallelism of DML, too. My simple (probably silly) question was, in INSERT SELECT,\\r\\n\\r\\n* If the target table has 10 partitions and the source table has 100 partitions, how would the user want to specify parameters?\\r\\n\\r\\n* If the source and target tables have the same number of partitions, and the user specified different values to parallel_workers and parallel_dml_workers, how many parallel workers would run?\\r\\n\\r\\n* What would the query plan be like? Something like below? Can we easily support this sort of nested thing?\\r\\n\\r\\nGather\\r\\n Workers Planned: <parallel_dml_workers>\\r\\n Insert\\r\\n Gather\\r\\n Workers Planned: <parallel_workers>\\r\\n Parallel Seq Scan\\r\\n\\r\\n\\r\\n> Which memory specific to partitions are you referring to here and does\\r\\n> that apply to the patch being discussed?\\r\\n\\r\\nRelation cache and catalog cache, which are not specific to partitions. This patch's current parallel safety check opens and closes all descendant partitions of the target table. That leaves those cache entries in CacheMemoryContext after the SQL statement ends. But as I said, we can consider it's not a serious problem in this case because the parallel DML would be executed in limited number of concurrent sessions. I just touched on the memory consumption issue for completeness in comparison with (3).\\r\\n\\r\\n\\r\\nRegards\\r\\nTakayuki Tsunakawa\\r\\n\\r\\n\",\n \"msg_date\": \"Mon, 18 Jan 2021 04:57:06 +0000\",\n \"msg_from\": \"\\\"tsunakawa.takay@fujitsu.com\\\" <tsunakawa.takay@fujitsu.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"RE: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"On Mon, Jan 18, 2021 at 10:27 AM tsunakawa.takay@fujitsu.com\\n<tsunakawa.takay@fujitsu.com> wrote:\\n>\\n> From: Amit Kapila <amit.kapila16@gmail.com>\\n> > We already allow users to specify the degree of parallelism for all\\n> > the parallel operations via guc's max_parallel_maintenance_workers,\\n> > max_parallel_workers_per_gather, then we have a reloption\\n> > parallel_workers and vacuum command has the parallel option where\\n> > users can specify the number of workers that can be used for\\n> > parallelism. The parallelism considers these as hints but decides\\n> > parallelism based on some other parameters like if there are that many\\n> > workers available, etc. Why the users would expect differently for\\n> > parallel DML?\\n>\\n> I agree that the user would want to specify the degree of parallelism of DML, too. My simple (probably silly) question was, in INSERT SELECT,\\n>\\n> * If the target table has 10 partitions and the source table has 100 partitions, how would the user want to specify parameters?\\n>\\n> * If the source and target tables have the same number of partitions, and the user specified different values to parallel_workers and parallel_dml_workers, how many parallel workers would run?\\n>\\n\\nGood question. I think if we choose to have a separate parameter for\\nDML, it can probably a boolean to just indicate whether to enable\\nparallel DML for a specified table and use the parallel_workers\\nspecified in the table used in SELECT.\\n\\n-- \\nWith Regards,\\nAmit Kapila.\\n\\n\\n\",\n \"msg_date\": \"Mon, 18 Jan 2021 14:03:55 +0530\",\n \"msg_from\": \"Amit Kapila <amit.kapila16@gmail.com>\",\n \"msg_from_op\": true,\n \"msg_subject\": \"Re: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"From: Amit Kapila <amit.kapila16@gmail.com>\\r\\n> Good question. I think if we choose to have a separate parameter for\\r\\n> DML, it can probably a boolean to just indicate whether to enable\\r\\n> parallel DML for a specified table and use the parallel_workers\\r\\n> specified in the table used in SELECT.\\r\\n\\r\\nAgreed.\\r\\n\\r\\n\\r\\nRegards\\r\\nTakayuki Tsunakawa\\r\\n\\r\\n\",\n \"msg_date\": \"Mon, 18 Jan 2021 08:37:26 +0000\",\n \"msg_from\": \"\\\"tsunakawa.takay@fujitsu.com\\\" <tsunakawa.takay@fujitsu.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"RE: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"> From: Amit Kapila <amit.kapila16@gmail.com>\\r\\n> > Good question. I think if we choose to have a separate parameter for\\r\\n> > DML, it can probably a boolean to just indicate whether to enable\\r\\n> > parallel DML for a specified table and use the parallel_workers\\r\\n> > specified in the table used in SELECT.\\r\\n> \\r\\n> Agreed.\\r\\n\\r\\nHi\\r\\n\\r\\nI have an issue about the parameter for DML.\\r\\n\\r\\nIf we define the parameter as a tableoption.\\r\\n\\r\\nFor a partitioned table, if we set the parent table's parallel_dml=on,\\r\\nand set one of its partition parallel_dml=off, it seems we can not divide the plan for the separate table.\\r\\n\\r\\nFor this case, should we just check the parent's parameter ?\\r\\n\\r\\nBest regards,\\r\\nhouzj\\r\\n\\n\\n\",\n \"msg_date\": \"Thu, 28 Jan 2021 11:30:43 +0000\",\n \"msg_from\": \"\\\"Hou, Zhijie\\\" <houzj.fnst@cn.fujitsu.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"RE: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"On Thu, Jan 28, 2021 at 5:00 PM Hou, Zhijie <houzj.fnst@cn.fujitsu.com> wrote:\\n>\\n> > From: Amit Kapila <amit.kapila16@gmail.com>\\n> > > Good question. I think if we choose to have a separate parameter for\\n> > > DML, it can probably a boolean to just indicate whether to enable\\n> > > parallel DML for a specified table and use the parallel_workers\\n> > > specified in the table used in SELECT.\\n> >\\n> > Agreed.\\n>\\n> Hi\\n>\\n> I have an issue about the parameter for DML.\\n>\\n> If we define the parameter as a tableoption.\\n>\\n> For a partitioned table, if we set the parent table's parallel_dml=on,\\n> and set one of its partition parallel_dml=off, it seems we can not divide the plan for the separate table.\\n>\\n> For this case, should we just check the parent's parameter ?\\n>\\n\\nI think so. IIUC, this means the Inserts where target table is parent\\ntable, those will just check the option of the parent table and then\\nignore the option value for child tables whereas we will consider\\nchildrel's option for Inserts where target table is one of the child\\ntable, right?\\n\\n-- \\nWith Regards,\\nAmit Kapila.\\n\\n\\n\",\n \"msg_date\": \"Thu, 28 Jan 2021 17:58:14 +0530\",\n \"msg_from\": \"Amit Kapila <amit.kapila16@gmail.com>\",\n \"msg_from_op\": true,\n \"msg_subject\": \"Re: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"> > Hi\\r\\n> >\\r\\n> > I have an issue about the parameter for DML.\\r\\n> >\\r\\n> > If we define the parameter as a tableoption.\\r\\n> >\\r\\n> > For a partitioned table, if we set the parent table's parallel_dml=on,\\r\\n> > and set one of its partition parallel_dml=off, it seems we can not divide\\r\\n> the plan for the separate table.\\r\\n> >\\r\\n> > For this case, should we just check the parent's parameter ?\\r\\n> >\\r\\n> \\r\\n> I think so. IIUC, this means the Inserts where target table is parent table,\\r\\n> those will just check the option of the parent table and then ignore the\\r\\n> option value for child tables whereas we will consider childrel's option\\r\\n> for Inserts where target table is one of the child table, right?\\r\\n> \\r\\n\\r\\nYes, I think we can just check the target table itself.\\r\\n\\r\\nBest regards,\\r\\nhouzj\\r\\n\\r\\n\\n\\n\",\n \"msg_date\": \"Fri, 29 Jan 2021 01:01:01 +0000\",\n \"msg_from\": \"\\\"Hou, Zhijie\\\" <houzj.fnst@cn.fujitsu.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"RE: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"Hi,\\r\\n\\r\\nAttatching v1 patches, introducing options which let user manually control whether to use parallel dml.\\r\\n\\r\\nAbout the patch:\\r\\n1. add a new guc option: enable_parallel_dml (boolean)\\r\\n2. add a new tableoption: parallel_dml (boolean)\\r\\n\\r\\n The default of both is off(false).\\r\\n\\r\\nUser can set enable_parallel_dml in postgresql.conf or seesion to enable parallel dml.\\r\\nIf user want to choose some specific to use parallel insert, they can set table.parallel_dml to on.\\r\\n\\r\\nSome attention：\\r\\n(1)\\r\\nCurrently if guc option enable_parallel_dml is set to on but table's parallel_dml is off,\\r\\nplanner still do not consider parallel for dml.\\r\\n\\r\\nIn this way, If user want to use parallel dml , they have to set enable_parallel_dml=on and set parallel_dml = on.\\r\\nIf someone dislike this, I think we can also let tableoption. parallel_dml's default value to on ,with this user only need\\r\\nto set enable_parallel_dml=on\\r\\n\\r\\n(2)\\r\\nFor the parallel_dml.\\r\\nIf target table is partitioned, and it's parallel_dml is set to on, planner will ignore\\r\\nThe option value of it's child. \\r\\nThis is beacause we can not divide dml plan to separate table, so we just check the target table itself.\\r\\n\\r\\n\\r\\nThoughts and comments are welcome.\\r\\n\\r\\nBest regards,\\r\\nhouzj\",\n \"msg_date\": \"Fri, 29 Jan 2021 06:43:14 +0000\",\n \"msg_from\": \"\\\"Hou, Zhijie\\\" <houzj.fnst@cn.fujitsu.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"RE: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"> Hi,\\r\\n> \\r\\n> Attatching v1 patches, introducing options which let user manually control\\r\\n> whether to use parallel dml.\\r\\n> \\r\\n> About the patch:\\r\\n> 1. add a new guc option: enable_parallel_dml (boolean) 2. add a new\\r\\n> tableoption: parallel_dml (boolean)\\r\\n> \\r\\n> The default of both is off(false).\\r\\n> \\r\\n> User can set enable_parallel_dml in postgresql.conf or seesion to enable\\r\\n> parallel dml.\\r\\n> If user want to choose some specific to use parallel insert, they can set\\r\\n> table.parallel_dml to on.\\r\\n> \\r\\n> Some attention：\\r\\n> (1)\\r\\n> Currently if guc option enable_parallel_dml is set to on but table's\\r\\n> parallel_dml is off, planner still do not consider parallel for dml.\\r\\n> \\r\\n> In this way, If user want to use parallel dml , they have to set\\r\\n> enable_parallel_dml=on and set parallel_dml = on.\\r\\n> If someone dislike this, I think we can also let tableoption. parallel_dml's\\r\\n> default value to on ,with this user only need to set enable_parallel_dml=on\\r\\n> \\r\\n> (2)\\r\\n> For the parallel_dml.\\r\\n> If target table is partitioned, and it's parallel_dml is set to on, planner\\r\\n> will ignore The option value of it's child.\\r\\n> This is beacause we can not divide dml plan to separate table, so we just\\r\\n> check the target table itself.\\r\\n> \\r\\n> \\r\\n> Thoughts and comments are welcome.\\r\\n\\r\\nThe patch is based on v13 patch(parallel insert select) in [1]\\r\\n\\r\\n[1] https://www.postgresql.org/message-id/CAJcOf-ejV8iU%2BYpuV4qbYEY-2vCG1QF2g3Gxn%3DZ%2BPyUH_5f84A%40mail.gmail.com\\r\\n\\r\\nBest regards,\\r\\nhouzj\\r\\n\\n\\n\",\n \"msg_date\": \"Fri, 29 Jan 2021 06:49:05 +0000\",\n \"msg_from\": \"\\\"Hou, Zhijie\\\" <houzj.fnst@cn.fujitsu.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"RE: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"On Fri, Jan 29, 2021 at 5:44 PM Hou, Zhijie <houzj.fnst@cn.fujitsu.com> wrote:\\n>\\n>\\n> Attatching v1 patches, introducing options which let user manually control whether to use parallel dml.\\n>\\n> About the patch:\\n> 1. add a new guc option: enable_parallel_dml (boolean)\\n> 2. add a new tableoption: parallel_dml (boolean)\\n>\\n> The default of both is off(false).\\n>\\n> User can set enable_parallel_dml in postgresql.conf or seesion to enable parallel dml.\\n> If user want to choose some specific to use parallel insert, they can set table.parallel_dml to on.\\n>\\n> Some attention：\\n> (1)\\n> Currently if guc option enable_parallel_dml is set to on but table's parallel_dml is off,\\n> planner still do not consider parallel for dml.\\n>\\n> In this way, If user want to use parallel dml , they have to set enable_parallel_dml=on and set parallel_dml = on.\\n> If someone dislike this, I think we can also let tableoption. parallel_dml's default value to on ,with this user only need\\n> to set enable_parallel_dml=on\\n>\\n> (2)\\n> For the parallel_dml.\\n> If target table is partitioned, and it's parallel_dml is set to on, planner will ignore\\n> The option value of it's child.\\n> This is beacause we can not divide dml plan to separate table, so we just check the target table itself.\\n>\\n>\\n> Thoughts and comments are welcome.\\n>\\n\\nPersonally, I think a table's \\\"parallel_dml\\\" option should be ON by default.\\nIt's annoying to have to separately enable it for each and every table\\nbeing used, when I think the need to turn it selectively OFF should be\\nfairly rare.\\nAnd I'm not sure that \\\"parallel_dml\\\" is the best name for the table\\noption - because it sort of implies parallel dml WILL be used - but\\nthat isn't true, it depends on other factors too.\\nSo I think (to be consistent with other table option naming) it would\\nhave to be something like \\\"parallel_dml_enabled\\\".\\nI'm still looking at the patches, but have so far noticed that there\\nare some issues in the documentation updates (grammatical issues and\\nconsistency issues with current documentation), and also some of the\\nexplanations are not clear. I guess I can address these separately.\\n\\nRegards,\\nGreg Nancarrow\\nFujitsu Australia\\n\\n\\n\",\n \"msg_date\": \"Mon, 1 Feb 2021 13:37:43 +1100\",\n \"msg_from\": \"Greg Nancarrow <gregn4422@gmail.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"Re: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"Hi greg,\\r\\n\\r\\nThanks for the review !\\r\\n\\r\\n> Personally, I think a table's \\\"parallel_dml\\\" option should be ON by default.\\r\\n> It's annoying to have to separately enable it for each and every table being\\r\\n> used, when I think the need to turn it selectively OFF should be fairly\\r\\n> rare.\\r\\n\\r\\nYes, I agreed.\\r\\nChanged.\\r\\n\\r\\n> And I'm not sure that \\\"parallel_dml\\\" is the best name for the table option\\r\\n> - because it sort of implies parallel dml WILL be used - but that isn't\\r\\n> true, it depends on other factors too.\\r\\n> So I think (to be consistent with other table option naming) it would have\\r\\n> to be something like \\\"parallel_dml_enabled\\\".\\r\\n\\r\\nAgreed.\\r\\nChanged to parallel_dml_enabled.\\r\\n\\r\\nAttatching v2 patch which addressed the comments above.\\r\\n\\r\\nSome further refactor:\\r\\n\\r\\nIntroducing a new function is_parallel_possible_for_modify() which decide whether to do safety check.\\r\\n\\r\\nIMO, It seems more readable to extract all the check that we can do before the safety-check and put them\\r\\nin the new function.\\r\\n\\r\\nPlease consider it for further review.\\r\\n\\r\\nBest regards,\\r\\nhouzj\",\n \"msg_date\": \"Mon, 1 Feb 2021 05:02:18 +0000\",\n \"msg_from\": \"\\\"Hou, Zhijie\\\" <houzj.fnst@cn.fujitsu.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"RE: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"On Mon, Feb 1, 2021 at 4:02 PM Hou, Zhijie <houzj.fnst@cn.fujitsu.com> wrote:\\n>\\n> Attatching v2 patch which addressed the comments above.\\n>\\n> Some further refactor:\\n>\\n> Introducing a new function is_parallel_possible_for_modify() which decide whether to do safety check.\\n>\\n> IMO, It seems more readable to extract all the check that we can do before the safety-check and put them\\n> in the new function.\\n>\\n> Please consider it for further review.\\n>\\n\\nI've updated your v2 patches and altered some comments and\\ndocumentation changes (but made no code changes) - please compare\\nagainst your v2 patches, and see whether you agree with the changes to\\nthe wording.\\nIn the documentation, you will also notice that in your V2 patch, it\\nsays that the \\\"parallel_dml_enabled\\\" table option defaults to false.\\nAs it actually defaults to true, I changed that in the documentation\\ntoo.\\n\\nRegards,\\nGreg Nancarrow\\nFujitsu Australia\",\n \"msg_date\": \"Tue, 2 Feb 2021 10:55:40 +1100\",\n \"msg_from\": \"Greg Nancarrow <gregn4422@gmail.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"Re: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"Hi,\\nFor v3_0003-reloption-parallel_dml-src.patch :\\n\\n+ * Check if parallel_dml_enabled is enabled for the target table,\\n+ * if not, skip the safety checks and return PARALLEL_UNSAFE.\\n\\nLooks like the return value is true / false. So the above comment should be\\nadjusted.\\n\\n+ if (!RelationGetParallelDML(rel, true))\\n+ {\\n+ table_close(rel, NoLock);\\n+ return false;\\n+ }\\n+\\n+ table_close(rel, NoLock);\\n\\nSince the rel would always be closed, it seems the return value\\nfrom RelationGetParallelDML() can be assigned to a variable, followed by\\ncall to table_close(), then the return statement.\\n\\nCheers\\n\\nOn Mon, Feb 1, 2021 at 3:56 PM Greg Nancarrow <gregn4422@gmail.com> wrote:\\n\\n> On Mon, Feb 1, 2021 at 4:02 PM Hou, Zhijie <houzj.fnst@cn.fujitsu.com>\\n> wrote:\\n> >\\n> > Attatching v2 patch which addressed the comments above.\\n> >\\n> > Some further refactor:\\n> >\\n> > Introducing a new function is_parallel_possible_for_modify() which\\n> decide whether to do safety check.\\n> >\\n> > IMO, It seems more readable to extract all the check that we can do\\n> before the safety-check and put them\\n> > in the new function.\\n> >\\n> > Please consider it for further review.\\n> >\\n>\\n> I've updated your v2 patches and altered some comments and\\n> documentation changes (but made no code changes) - please compare\\n> against your v2 patches, and see whether you agree with the changes to\\n> the wording.\\n> In the documentation, you will also notice that in your V2 patch, it\\n> says that the \\\"parallel_dml_enabled\\\" table option defaults to false.\\n> As it actually defaults to true, I changed that in the documentation\\n> too.\\n>\\n> Regards,\\n> Greg Nancarrow\\n> Fujitsu Australia\\n>\\n\\nHi,For v3_0003-reloption-parallel_dml-src.patch :+ * Check if parallel_dml_enabled is enabled for the target table,+ * if not, skip the safety checks and return PARALLEL_UNSAFE.Looks like the return value is true / false. So the above comment should be adjusted.+ if (!RelationGetParallelDML(rel, true))+ {+ table_close(rel, NoLock);+ return false;+ }++ table_close(rel, NoLock);Since the rel would always be closed, it seems the return value from RelationGetParallelDML() can be assigned to a variable, followed by call to table_close(), then the return statement.CheersOn Mon, Feb 1, 2021 at 3:56 PM Greg Nancarrow <gregn4422@gmail.com> wrote:On Mon, Feb 1, 2021 at 4:02 PM Hou, Zhijie <houzj.fnst@cn.fujitsu.com> wrote:\\n>\\n> Attatching v2 patch which addressed the comments above.\\n>\\n> Some further refactor:\\n>\\n> Introducing a new function is_parallel_possible_for_modify() which decide whether to do safety check.\\n>\\n> IMO, It seems more readable to extract all the check that we can do before the safety-check and put them\\n> in the new function.\\n>\\n> Please consider it for further review.\\n>\\n\\nI've updated your v2 patches and altered some comments and\\ndocumentation changes (but made no code changes) - please compare\\nagainst your v2 patches, and see whether you agree with the changes to\\nthe wording.\\nIn the documentation, you will also notice that in your V2 patch, it\\nsays that the \\\"parallel_dml_enabled\\\" table option defaults to false.\\nAs it actually defaults to true, I changed that in the documentation\\ntoo.\\n\\nRegards,\\nGreg Nancarrow\\nFujitsu Australia\",\n \"msg_date\": \"Mon, 1 Feb 2021 16:04:11 -0800\",\n \"msg_from\": \"Zhihong Yu <zyu@yugabyte.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"Re: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"> > IMO, It seems more readable to extract all the check that we can do\\r\\n> > before the safety-check and put them in the new function.\\r\\n> >\\r\\n> > Please consider it for further review.\\r\\n> >\\r\\n> \\r\\n> I've updated your v2 patches and altered some comments and documentation\\r\\n> changes (but made no code changes) - please compare against your v2 patches,\\r\\n> and see whether you agree with the changes to the wording.\\r\\n> In the documentation, you will also notice that in your V2 patch, it says\\r\\n> that the \\\"parallel_dml_enabled\\\" table option defaults to false.\\r\\n> As it actually defaults to true, I changed that in the documentation too.\\r\\n\\r\\nHi greg,\\r\\n\\r\\nThanks a lot for the document update, LGTM.\\r\\n\\r\\nAttaching v4 patches with the changes:\\r\\n * fix some typos in the code.\\r\\n * store partitioned reloption in a separate struct PartitionedOptions, \\r\\n making it more standard and easier to expand in the future.\\r\\n\\r\\nPlease consider it for further review.\\r\\n\\r\\nBest regards,\\r\\nHouzj\",\n \"msg_date\": \"Tue, 2 Feb 2021 01:54:37 +0000\",\n \"msg_from\": \"\\\"Hou, Zhijie\\\" <houzj.fnst@cn.fujitsu.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"RE: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"Hi,\\r\\n\\r\\nAttaching v5 patches with the changes:\\r\\n * rebase the code on the greg's latest parallel insert patch\\r\\n * fix some code style.\\r\\n\\r\\nPlease consider it for further review.\\r\\n\\r\\nBest regards,\\r\\nHouzj\",\n \"msg_date\": \"Wed, 3 Feb 2021 01:01:05 +0000\",\n \"msg_from\": \"\\\"Hou, Zhijie\\\" <houzj.fnst@cn.fujitsu.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"RE: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"> For v3_0003-reloption-parallel_dml-src.patch :\\r\\n> + table_close(rel, NoLock);\\r\\n\\r\\n> Since the rel would always be closed, it seems the return value from RelationGetParallelDML() can be assigned to a variable, followed by call to table_close(), then the return statement.\\r\\n\\r\\nThanks for the comment. Fixed in the latest patch.\\r\\n\\r\\nBest regards,\\r\\nhouzj\\r\\n\\n\\n\",\n \"msg_date\": \"Wed, 3 Feb 2021 01:02:45 +0000\",\n \"msg_from\": \"\\\"Hou, Zhijie\\\" <houzj.fnst@cn.fujitsu.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"RE: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"Hi,\\r\\n\\r\\nI notice the comment 5) about is_parallel_possible_for_modify() seems to be inaccurate in clauses.c.\\r\\n\\t* 5) the reloption parallel_dml_enabled is not set for the target table\\r\\n\\r\\nBecause you have set parallel_dml_enabled to 'true' as default. \\r\\n\\t{\\r\\n\\t\\t{\\r\\n\\t\\t\\t\\\"parallel_dml_enabled\\\",\\r\\n\\t\\t\\t\\\"Enables \\\\\\\"parallel dml\\\\\\\" feature for this table\\\",\\r\\n\\t\\t\\tRELOPT_KIND_HEAP | RELOPT_KIND_PARTITIONED,\\r\\n\\t\\t\\tShareUpdateExclusiveLock\\r\\n\\t\\t},\\r\\n\\t\\ttrue\\r\\n\\t}\\r\\n\\r\\nSo even user doesn't set parallel_dml_enabled explicit, its value is 'on', Parallel is also possible for this rel(when enable_parallel_dml is on). \\r\\n\\r\\nMaybe we can just comment like this or a better one you'd like if you agree with above:\\r\\n\\t* 5) the reloption parallel_dml_enabled is set to off\\r\\n\\r\\nRegards\\r\\nHuang\\r\\n\\r\\n> -----Original Message-----\\r\\n> From: Hou, Zhijie <houzj.fnst@cn.fujitsu.com>\\r\\n> Sent: Wednesday, February 3, 2021 9:01 AM\\r\\n> To: Greg Nancarrow <gregn4422@gmail.com>\\r\\n> Cc: Amit Kapila <amit.kapila16@gmail.com>; PostgreSQL Hackers\\r\\n> <pgsql-hackers@lists.postgresql.org>; vignesh C <vignesh21@gmail.com>;\\r\\n> Amit Langote <amitlangote09@gmail.com>; David Rowley\\r\\n> <dgrowleyml@gmail.com>; Tom Lane <tgl@sss.pgh.pa.us>; Tsunakawa,\\r\\n> Takayuki/綱川 貴之 <tsunakawa.takay@fujitsu.com>\\r\\n> Subject: RE: Determine parallel-safety of partition relations for Inserts\\r\\n> \\r\\n> Hi,\\r\\n> \\r\\n> Attaching v5 patches with the changes:\\r\\n> * rebase the code on the greg's latest parallel insert patch\\r\\n> * fix some code style.\\r\\n> \\r\\n> Please consider it for further review.\\r\\n> \\r\\n> Best regards,\\r\\n> Houzj\\r\\n> \\r\\n> \\r\\n> \\r\\n> \\r\\n\\r\\n\\n\\n\",\n \"msg_date\": \"Fri, 5 Feb 2021 02:54:32 +0000\",\n \"msg_from\": \"\\\"Huang, Qiuyan\\\" <huangqy.fnst@cn.fujitsu.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"RE: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"Hi,\\r\\n\\r\\nAttaching v6 patches with the changes:\\r\\n * rebase the code on the greg's latest parallel insert patch.\\r\\n * fix some code comment.\\r\\n * add some test to cover the partitioned table.\\r\\n\\r\\nPlease consider it for further review.\\r\\n\\r\\nBest regards,\\r\\nHouzj\",\n \"msg_date\": \"Thu, 18 Feb 2021 03:24:48 +0000\",\n \"msg_from\": \"\\\"Hou, Zhijie\\\" <houzj.fnst@cn.fujitsu.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"RE: Determine parallel-safety of partition relations for Inserts\"\n },\n {\n \"msg_contents\": \"Hi,\\r\\n \\r\\nAttaching v7 patches with the changes:\\r\\n * rebase the code on the greg's latest parallel insert patch.\\r\\n\\r\\n Please consider it for further review.\\r\\n \\r\\n Best regards,\\r\\n houzj\",\n \"msg_date\": \"Mon, 22 Feb 2021 03:27:34 +0000\",\n \"msg_from\": \"\\\"houzj.fnst@fujitsu.com\\\" <houzj.fnst@fujitsu.com>\",\n \"msg_from_op\": false,\n \"msg_subject\": \"RE: Determine parallel-safety of partition relations for Inserts\"\n }\n]","isConversational":false}}}],"truncated":false,"partial":false},"paginationData":{"pageIndex":514,"numItemsPerPage":100,"numTotalItems":58371,"offset":51400,"length":100}},"jwt":"eyJhbGciOiJFZERTQSJ9.eyJyZWFkIjp0cnVlLCJwZXJtaXNzaW9ucyI6eyJyZXBvLmNvbnRlbnQucmVhZCI6dHJ1ZX0sImlhdCI6MTc4MDg0Njg0Mywic3ViIjoiL2RhdGFzZXRzL3dhbnNoZW5sL3Bnc3FsLWhhY2tlcnMiLCJleHAiOjE3ODA4NTA0NDMsImlzcyI6Imh0dHBzOi8vaHVnZ2luZ2ZhY2UuY28ifQ.tmBj3OBMgO1tRB1N-6Tdu43xZ9jHla_V_q73l3Ggw6KSFP0SBVUbZW5C0ZWYXEmsy9KUOkv0VQ7tfgT-MyCYDA","displayUrls":true,"splitSizeSummaries":[{"config":"default","split":"train","numRows":58371,"numBytesParquet":368254123}]},"discussionsStats":{"closed":0,"open":0,"total":0},"fullWidth":true,"hasGatedAccess":true,"hasFullAccess":true,"isEmbedded":false,"savedQueries":{"community":[],"user":[]}}">
threads listlengths 1 2.99k |
|---|
[
{
"msg_contents": "Hi all,\n\nI've started this separate thread from [1] for discussing the general\nAPI design of index vacuum.\n\nSummary:\n\n* Call ambulkdelete and amvacuumcleanup even when INDEX_CLEANUP is\nfalse, and leave it to the index AM whether or not skip them.\n* Add a new index AM API amvacuumstra... |
[
{
"msg_contents": "Hi all,\n\nWhile looking at a patch from David, I have noticed $subject:\nhttps://www.postgresql.org/message-id/CAApHDvpgB+vxk=W6OPKidwzZEo6kniFQidNoMzR8P4ROtyky2w@mail.gmail.com\n\nadminpack and old_snapshot have no need for those references as they\ndon't use libpq. Any objections to a sma... |
[
{
"msg_contents": "Hi\n\nI am playing with clipboard on Linux, and I found a way, how to redirect\npsql output to clipboard via wl-copy or xclip and then to Libre Office. Now\nit looks so best format is tsv\n\nselect * from pg_database \\g (format=tsv) | wl-paste -t\napplication/x-libreoffice-tsvc\n\nImplementa... |
[
{
"msg_contents": "Hi all\n\nI want to share an idea I've looked at a few times where I've run into\nsituations where logical slots were inadvertently dropped, or where it\nbecame necessary to decode changes in the past on a slot.\n\nAs most of you will know you can't just create a logical slot in the past.\nEv... |
[
{
"msg_contents": "Hi all\n\nI recently wrote a utility that adds a $node->gdb_backends() method to\nPostgresNode instances - figured I'd share it here in case anyone finds it\nuseful, or wants to adopt it into the features of the TAP tools.\n\nThis function provides a one-line way to dump stacks for all runnin... |
[
{
"msg_contents": "I'm trying to store connection to postgres_fdw in the database I want to\nbe able to store the full breadth of connection styles and all the\ndifferent types of connections that libpq supports. But having some\ntroubles.\n\nPostgres_fdw wants options passed into CREATE SERVER, all broken ou... |
[
{
"msg_contents": "Currently, session poolers operating in transaction mode need to send\na \"server_reset_query\" which is mostly DISCARD ALL.\n\nIt seems strange to me that we put this work onto the pooler, forcing\npoolers to repeatedly issue the same command, at some cost in\nperformance. Measuring the over... |
[
{
"msg_contents": "Hi,\nI searched for Postgres support in Apache Spark.\nI found Spark doc related to JDBC.\n\nI wonder if the community is aware of Spark connector for Postgres\n(hopefully open source) where predicate involving jsonb columns can be\npushed down.\n\nYour input is appreciated.\n\nThanks\n\nHi,I... |
[
{
"msg_contents": "\nHello,\n\nI tried:\n\n psql> SELECT TIMESTAMP '2020-12-23Z19:28:45';\n\nThe result of which is:\n\n 2020-12-23 00:00:00\n\nThis is disappointing. Ok, my fault, I should have written TIMESTAMPTZ, or \nuse T instead of Z, or whatever.\n\nAnyway, is there a rational for this behavior? I woul... |
[
{
"msg_contents": "Hi, hackers\r\n\r\nWhen I use logical stream replication on partition table, I find that if we create a new\r\npartitions after the subscription on subscriber, the records in new partitions cannot be\r\nshipped to the subscriber.\r\n\r\nHere is an example:\r\n\r\n1. Create a view to check th... |
[
{
"msg_contents": "Hello, hackers!\n\n2021-01 commitfest will start from the new year, in a week. I'm happy\nto volunteer to be the CFM for this one.\n\nIt's time to register your patch in the commitfest, if not yet.\n\nIf you already have a patch in the commitfest, update its status and\nmake sure it still app... |
[
{
"msg_contents": "Hello.\n\nI'm occasionally annoyed by the behavior of pg_waldump that it shows\nall records in the specified segment before entering waiting state.\n\ntail -f shows only the last few dozen lines at the start. This\nbehavior I think is useful also for pg_waldump -f.\n\nIt was as simple than e... |
[
{
"msg_contents": "Add key management system\n\nThis adds a key management system that stores (currently) two data\nencryption keys of length 128, 192, or 256 bits. The data keys are\nAES256 encrypted using a key encryption key, and validated via GCM\ncipher mode. A command to obtain the key encryption key mu... |
[
{
"msg_contents": "Hi,\n\nWhen running write heavy transactional workloads I've many times\nobserved that one needs to run the benchmarks for quite a while till\nthey get to their steady state performance. The most significant reason\nfor that is that initially WAL files will not get recycled, but need to\nbe f... |
[
{
"msg_contents": "Add pg_alterckey utility to change the cluster key\n\nThis can change the key that encrypts the data encryption keys used for\ncluster file encryption.\n\nDiscussion: https://postgr.es/m/20201202213814.GG20285@momjian.us\n\nBackpatch-through: master\n\nBranch\n------\nmaster\n\nDetails\n-----... |
[
{
"msg_contents": "\n\n /*\n * Return a copy of heap tuple representing the contents of the slot.\nThe\n * copy needs to be palloc'd in the current memory context. The slot\n * itself is expected to remain unaffected. It is *not* expected to\nhave\n * meaningful \"system columns\" in the copy. The copy... |
[
{
"msg_contents": "Hi\n\nmaybe interesting project for decreasing overhead of JIT\n\nhttps://developers.redhat.com/blog/2020/01/20/mir-a-lightweight-jit-compiler-project/\n\nRegards\n\nPavel\n\nHimaybe interesting project for decreasing overhead of JIThttps://developers.redhat.com/blog/2020/01/20/mir-a-lightwei... |
[
{
"msg_contents": "Hi,\n\nWhen developing extensions, I find myself spending time on manually checking\nif my update scripts (invoked via ALTER EXTENSION ... UPDATE) gives the same result as\nCREATE EXTENSION ... would do if installing the latest version from scratch without any previous version.\n\nI thought i... |
[
{
"msg_contents": "vacuum_cost_page_miss has a default value of 10, while\nvacuum_cost_page_dirty has a default value of 20. This has been the\ncase since cost-based delays were introduced by commit f425b605f4e\nback in 2004. The obvious implication is that dirtying a page is on\naverage only twice as expensive... |
[
{
"msg_contents": "https://postgr.es/m/20201031163518.GB4039133@rfd.leadboat.com gave $SUBJECT as\none of the constituent projects for changing the public schema default ACL.\nThis feature stands on its own, hence the new thread. I showed syntax \"ALTER\nSCHEMA public OWNER TO DATABASE_OWNER\" and anticipated ... |
[
{
"msg_contents": "Hi,\n\nI would like to propose a small patch to the JIT machinery which makes \nthe IR code generation lazy. The reason for postponing the generation of \nthe IR code is that with partitions we get an explosion in the number of \nJIT functions generated as many child tables are involved, each... |
[
{
"msg_contents": "I was surprised to find that this doesn't error:\n\n=> select 100a;\n a\n-----\n 100\n\nI suspect this and similar cases used to error before aliases without AS \nwere introduced. But now this seems possibly problematic. Should we \ntry to handle this better?\n\n\n",
"msg_date": "Mon... |
[
{
"msg_contents": "Back in 2001 we decided to prefer putenv() over setenv() because the\nlatter wasn't in POSIX (SUSv2) at the time. That decision has been\novertaken by events: more recent editions of POSIX not only provide\nsetenv() but recommend it over putenv(). So I think it's time to\nchange our policy ... |
[
{
"msg_contents": "Hi all,\n\nBuildfarm member gombessa just had an interesting failure:\nhttps://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=gombessa&dt=2020-12-29%2000%3A16%3A49\n Seq Scan on int8_tbl i8 (cost=N.N..N.N rows=N width=N) (actual\n time=N.N..N.N rows=N loops=N)\n Planning Time: N.N ms\n- ... |
[
{
"msg_contents": "https://postgr.es/m/20201031163518.GB4039133@rfd.leadboat.com gave $SUBJECT as\none of the constituent projects for changing the public schema default ACL.\nThis ended up being simple. Attached. I chose to omit the \"ALTER SCHEMA\npublic OWNER TO\" when the owner is the bootstrap superuser,... |
[
{
"msg_contents": "Hey, hackers,\n\nAs part of building Postgres for a managed environment in Google Cloud SQL,\nwe've made various small changes to permission checks to carefully limit what\ncustomers have access to.\n\nWe were making some changes in src/backend/commands/user.c and noticed some\nclunky logic r... |
[
{
"msg_contents": "Hi,\n\nThis started as a reply to\nhttps://postgr.es/m/CAH2-WzkbCvgKrmw%2Bf%2B1hwgXhmiv%2BUNRihotALXftUiNr%3D3VUKA%40mail.gmail.com\nbut after typing for a while I decided that it's large and unrelated\nenough to be better handled as a new thread.\n\n\nOn 2020-11-27 11:08:49 -0800, Peter Geog... |
[
{
"msg_contents": "Hi,\n\nPer request, I'd like to see about surfacing $Subject to SQL.\n\nBest,\nDavid.\n-- \nDavid Fetter <david(at)fetter(dot)org> http://fetter.org/\nPhone: +1 415 235 3778\n\nRemember to vote!\nConsider donating to Postgres: http://www.postgresql.org/about/donate",
"msg_date": "Wed, 30 ... |
[
{
"msg_contents": "The first file of Amit's patch can not only re-range the code, but also fix a hidden bug.\r\nTo make it easy to see, I attach another patch.\r\n\"RelationIdGetRelation\" will increase ref on owner->relrefarr, without \"RelationClose\", the owner->relrefarr will enlarge and re-hash.\r\nW... |
[
{
"msg_contents": "I broke $SUBJECT again today by removing regress_putenv from\nregress.so. I don't really want to put it back, but that means\nwe need a way to drop the corresponding SQL function from the\nold version's regression database before attempting an upgrade.\n\nSince I'd just seen Noah's commit 52... |
[
{
"msg_contents": "Hi,\n\nA question from a colleague made me wonder if there are scenarios where\ntwo subsequent crashes could lead to wrong WAL to be applied.\n\nImagine the following scenario\n[ xlog page 1 ][ xlog page 2 ][ xlog page 3 ][ xlog page 4 ]\n ^flush ^wri... |
[
{
"msg_contents": "I now understand the wisdom of moving the remaining hex functions to\n/common. I know someone already suggested that, and the attached patch\ndoes this.\n\nI will add the attached patch to the commitfest so I can get cfbot\ntesting.\n\n-- \n Bruce Momjian <bruce@momjian.us> https://... |
[
{
"msg_contents": "Please find attached :)\n\nThis could probably use a lot of filling in, but having it in the\nactual documentation beats needing to know folklore even to know\nthat the capability is there.\n\nBest,\nDavid.\n-- \nDavid Fetter <david(at)fetter(dot)org> http://fetter.org/\nPhone: +1 415 235 377... |
[
{
"msg_contents": "Hi,\n\nin today's world, some DBAs have no superuser rights, but we can\ndelegate them additional powers like CREATEDB or the pg_monitor default\nrole etc. Usually, the DBA can also view the database logs, either via\nshell access or some web interface.\n\nOne thing that I personally find lac... |
[
{
"msg_contents": "Here's a rebase of Simon/Pavan's MERGE patch to current sources. I\ncleaned up some minor things in it, but aside from rebasing, it's pretty\nmuch their work (even the commit message is Simon's).\n\nAdding to commitfest.\n\n-- \n�lvaro Herrera",
"msg_date": "Thu, 31 Dec 2020 10:47:36 -... |
[
{
"msg_contents": "Hello,\n\nfinally I had some time to revisit patch and all comments from\nhttps://www.postgresql.org/message-id/CAFp7QwqMGEi4OyyaLEK9DR0%2BE%2BoK3UtA4bEjDVCa4bNkwUY2PQ%40mail.gmail.com\nand I have prepared simple version of COPY command progress reporting.\n\nTo keep the patch small as possib... |
[
{
"msg_contents": "I have completed the key management patch with tests created by Stephen\nFrost. Original patch by Masahiko Sawada. It requires the hex\nreorganization patch first. The key patch is now 2.1MB because of the\ntests, so attaching it here seems unwise:\n\n\thttps://github.com/postgres/postgres... |
[
{
"msg_contents": "I have long maintained permissions in my databases by having a script which\nassigns all the permissions. I have tended to start with invocations\nsomething like this:\n\nREVOKE ALL ON ALL TABLES IN SCHEMA ... FROM ...;\n\n... where the appropriate schemas and roles are listed. This is intend... |
[
{
"msg_contents": "Hi,\n\nHappy new year to all!\n\nThe January Commitfest for PG14 is now in progress! I'm happy to\nvolunteer to manage it.\n\nCurrent state for the Commitfest is:\n\nNeeds review: 188\nWaiting on Author: 28\nReady for Committer: 22\nCommitted: 19\nWithdrawn: 3\nTotal: 260\n\nThe number of pat... |
[
{
"msg_contents": "Hi,\n\nIn an effort to speed up bulk data loading/transforming I noticed that \nconsiderable time is spent in the relation extension lock. I know there \nare already many other efforts to increase the chances of using bulk \nloading [1], [2], [3], [4], efforts to make loading more parallel [5... |
[
{
"msg_contents": "Hi,\n\nI noticed -k/--data-checksums is currently in the less commonly used\noptions part of the initdb --help output:\n\n|Less commonly used options:\n| -d, --debug generate lots of debugging output\n| -k, --data-checksums use data page checksums\n\nI think enough people... |
[
{
"msg_contents": "Hello, hackers.\n\nWorking on some stuff, I realized I do not understand why\nlatestRemovedXid|cuteoff_xid (in different types of WAL records) are\nsent every time they appear on the primary side.\n\nlatestRemovedXid|cuteoff_xid is used to call\nResolveRecoveryConflictWithSnapshot and cancel ... |
[
{
"msg_contents": "Hi hackers,\n\nI just learned about a feature called \"LET clause\".\n\nIt's not part of the SQL standard, but it's supported by Oracle [1], Couchbase [2] and AsterixDB [3].\n\nI searched the pgsql-hackers archives and couldn't find any matches on \"LET clause\",\nso I thought I should share ... |
[
{
"msg_contents": "Hi, all,\n\nWhen using 'create index concurrently', we use a MVCC snaptshot, we should\nnot see tuples that are HEAPTUPLE_INSERT_IN_PROGRESS. I think we should\nreport an ERROR in this case. But, we report WARNING now.\n\nIn the early days, it indeed reports ERROR, but the behaviour is change... |
[
{
"msg_contents": "Hi;\n\n In a situation where pg_rewind gets an error due to a missing wall, I\nhave set restore_command so that the needed wals can be read from the\narchive (I don't want to manually copy the wal files), but I see it doesn't\nwork. What am I missing? Is restore_command not really workin... |
[
{
"msg_contents": "Over the holiday break I've been fooling with some regex performance\nimprovements. I don't have anything ready to show yet in that line,\nbut I was feeling the need for more-thorough test coverage, so I set\nto work on something that's been in the back of my mind for a long\ntime: we need t... |
[
{
"msg_contents": "I notice that heapdesc.c outputs the field latestRemovedXid as\n\"remxid\". But why? What sense is there in changing the name for output\nby tools like pg_waldump, which are intrinsically internals focussed?\n\nDoes anyone have any objections to my changing the details within\nheapdesc.c on m... |
[
{
"msg_contents": "Logical replication on partition table is very slow and CPU is 99%.\r\n\r\n\r\n\r\n\r\nTo reproduce this problem,\r\n\r\n\r\n\r\n\r\npg1 for publish\r\n\r\npgbench -i -s 1000\r\n\r\ncreate table pgbench_accounts_copy(aid integer, bid integer, abalance integer, filler character(84)) partition ... |
[
{
"msg_contents": "I just learned from a pg friend, it's possible to set your own made up setting_name using set_config(),\nif the setting_name contains at least one dot \".\".\n\nI didn't know about this and have always thought it's not possible,\ndue to early in my career having run into the error message\nyo... |
[
{
"msg_contents": "For example:\n\n$ python3.5 -c \"import pg; db=pg.DB(); q = db.query(\\\"SET log_parameter_max_length_on_error=-1;\\\"); db.prepare('p', 'SELECT \\$1::smallint'); db.query_prepared('p',66666);\"\n2021-01-03 02:21:04.547 CST [20157] ERROR: value \"66666\" is out of range for type smallint\n20... |
[
{
"msg_contents": "(Amusingly I only found this after discovering that Windows Calculator\nhas a similar bug which causes it to crash if you try to raise a\nnumber to the power INT_MIN.)\n\nOn my machine, numeric_power() loses all precision if the exponent is\nINT_MIN, though the actual failure mode might well ... |
[
{
"msg_contents": "Our developer FAQ mentions context diffs in several places for their\nimproved readability:\n\n\thttps://wiki.postgresql.org/wiki/Developer_FAQ\n\nI also know git can product context diff as _output_. However, there\nare several negatives to context vs unified diffs in our workflow:\n\n* \"... |
[
{
"msg_contents": "PSA a trivial patch to correct a comment typo.\n\n----\nKind Regards,\nPeter Smith\nFujitsu Australia.",
"msg_date": "Tue, 5 Jan 2021 11:06:24 +1100",
"msg_from": "Peter Smith <smithpb2250@gmail.com>",
"msg_from_op": true,
"msg_subject": "Fix typo in comment"
},
{
"msg... |
[
{
"msg_contents": "Hi.\n\nI wanted to add a new commitfest entry but although I am logged in I\ndo not see any \"Add\" button either in the current or future CF.\n\nhttps://commitfest.postgresql.org/31/\nhttps://commitfest.postgresql.org/32/\n\nHow to add?\n\n----\nKind Regards,\nPeter Smith.\nFujitsu Australia... |
[
{
"msg_contents": "Hi\n\nI found a possible typo in the comment of ReorderBufferProcessTXN().\n\n* relmapper has no \"historic\" view, in contrast to normal\n* the normal catalog during decoding. Thus repeated\n\nIs there an extra ‘normal’ in the comment ?\n\nin contrast to normal the normal catalog during dec... |
[
{
"msg_contents": "Hi,\n\nMy new favourite CI is Cirrus CI, because it has 4 operating systems,\ngenerous enough quotas to handle 250+ branches in a single account,\nand public build/test log URLs. I flipped cfbot.cputube.org (mostly)\nover to that and it seems to work well so far -- fingers crossed.\nI've als... |
[
{
"msg_contents": "While reviewing logical decoding of 2PC xacts work, I noticed that we\nneed $SUBJECT [1]. Commit 1eb6d6527a [2] allowed to track replica\norigin replay progress for 2PC but it was not complete. It misses to\nproperly track the progress for rollback prepared especially it missed\nto update the... |
[
{
"msg_contents": "Hi,\n\nIt seems like this part of the documentation was not updated after changing\nthe default value of recovery_target_timeline to latest in v12.\n\n\"The default behavior of recovery is to recover along the same timeline\nthat was current when the base backup was taken. If you wish to rec... |
[
{
"msg_contents": "Hello,\n\nThe PGresult structure returned on COPY ... FROM provides a handful of\ninformation (https://www.postgresql.org/docs/13/libpq-copy.html):\nnumber of fields expected and whether they are binary or not.\n\nHowever it doesn't provide the types expected. For binary copy the\nexact type ... |
[
{
"msg_contents": "The streaming replication protocol docs[1] say (for START_REPLICATION):\n\n\"If the client requests streaming at exactly the end of an old\ntimeline, the server responds immediately with CommandComplete without\nentering COPY mode.\"\n\nIs there a reason we do that? It seems like enough of an... |
[
{
"msg_contents": "hi,\n\n\ni updated our ci pipeline to the latest 14-devel build from the postgresql apt repository:\n\n\nPostgreSQL 14devel (Debian 14~~devel~20210105.1140-1~285.gitbc43b7c.pgdg110+1) on x86_64-pc-linux-gnu, compiled by gcc (Debian 10.2.1-3) 10.2.1 20201224, 64-bit\n\n\ni found, that the beha... |
[
{
"msg_contents": "Hey, hackers!\n\nI've created a patch to better support referential integrity constraints when\nusing composite primary and foreign keys. This patch allows creating a foreign\nkey using the syntax:\n\n FOREIGN KEY (tenant_id, fk_id) REFERENCES fktable ON DELETE SET NULL (fk_id)\n\nwhich mean... |
[
{
"msg_contents": "Hi,\n\nDue to a debug ereport I just noticed that worker.c's\nslot_store_error_callback is doing something quite dangerous:\n\nstatic void\nslot_store_error_callback(void *arg)\n{\n\tSlotErrCallbackArg *errarg = (SlotErrCallbackArg *) arg;\n\tLogicalRepRelMapEntry *rel;\n\tchar\t *remotetyp... |
[
{
"msg_contents": "Hello,\n\n\nWhile I'm reviewing patches for parallel operations, it occured to me that we may need to correct the words and enrich content in \"15. Parallel Query\". I don't have a concrete image of how the content should be yet, and I think the documentation may be improved after the featur... |
[
{
"msg_contents": "Hi,\n\n\nThis thread came from another thread about wal_level [1].\n\nMainly from backup management tools perspective\nsuch as pgBackRest, EDB's BART and pg_probackup,\nit seems worth talking about a way comprehensively\nto trace and recognize wal_level changes for various purposes and values... |
[
{
"msg_contents": "Hi ALL,\n\nSnip from CreateCheckPoint():\n--\n/*\n * Acquire CheckpointLock to ensure only one checkpoint happens at a time.\n * (This is just pro forma, since in the present system structure there is\n * only one process that is allowed to issue checkpoints at any given\n * time.)\n */\nLWLo... |
[
{
"msg_contents": "to be clear, that was existing code within our codebase, it was used as a very bad alternative to an if/else clause. think of it as an tenery operator. thats just straight up bad code as tom said. but even if it is, its a change and i wanted to check, if its intended. so thank you for you tim... |
[
{
"msg_contents": "Hi,\n\nThis feature adds schema option while creating publication. Users will\nbe able to specify one or more schemas while creating publication,\nwhen the user specifies schema option, then the data changes for the\ntables present in the schema specified by the user will be replicated\nto th... |
[
{
"msg_contents": "Hi hackers,\n\nMy learn-me-some-operating-system-hacking project for the holidays was\nto add O_DSYNC to FreeBSD 13 (due out end of Q1ish). I was motivated\nby my project to port some of Andres's Linux-only PostgreSQL AIO stuff\nto POSIX interfaces, where you need O_DSYNC to initiate the\nas... |
[
{
"msg_contents": "pg_dump can generate invalid SQL syntax if an aclitem in pg_init_privs\ncontains a GRANT OPTION (an asterisk in the aclitemout() representation).\nAlso, pg_dump uses REVOKE GRANT OPTION FOR where it needs plain REVOKE. For\nmore details, see the log message in the attached patch. I recommen... |
[
{
"msg_contents": "Hi,\n\nin expanded output, psql does not print the title if there are zero\nresults. In regular output, it prints the title (\"foo\") no matter how\nmany rows:\n\npostgres=# \\pset title foo\nTitle is \"foo\".\npostgres=# SELECT 1 WHERE 1=1;\n foo\n ?column? \n----------\n 1\n(1 row)... |
[
{
"msg_contents": "Hi,\n\nSince pg_test_fsync is supposed to simulate some aspects of\nPostgreSQL's wal_sync_method settings, I think it should be updated to\nuse the same system calls (which changed in v12). That's mostly on\nprinciple, though in practice, on one system I've managed to see a\nsmall measurable... |
[
{
"msg_contents": "In doc/src/sgml/func.sgml description of SHOW command use\n\"<acronym>SQL</acronym>\", while SET command description the same\nsection does not use \"<acronym>\". Shouldn't the description of SET use\n\"<acronym>\" for \"SQL\" as well? Patch attached.\n\nBest regards,\n--\nTatsuo Ishii\nSRA O... |
[
{
"msg_contents": "Hi hackers.\n\nWhen I using a Non-Exclusive Low-Level Backup on windows, \"invalid data in file backup_label\" will be shown.\n\nThe code are listed below\n\n\tif (fscanf(lfp, \"START WAL LOCATION: %X/%X (file %08X%16s)%c\",\n\t\t\t &hi, &lo, &tli_from_walseg, startxlogfilename, &ch) != 5 |... |
[
{
"msg_contents": "Hi,\n\nI noticed that RemoveXlogFile() has this code:\n\n /*\n * Before deleting the file, see if it can be recycled as a future log\n * segment. Only recycle normal files, pg_standby for example can create\n * symbolic links pointing to a separate archive direc... |
[
{
"msg_contents": "Hi,\n\nWhile providing thoughts on the design in [1], I found a strange\nbehaviour with the $subject. The use case is shown below as a sequence\nof steps that need to be run on publisher and subscriber to arrive at\nthe strange behaviour. In step 5, the table is dropped from the\npublication... |
[
{
"msg_contents": "Commit 5ee2197767 (about 4 years old) introduced the error:\n\n \"IDENTIFY_SYSTEM has not been run before START_REPLICATION\"\n\nBut it seems like running START_REPLICATION first works (at least in\nthe simple case).\n\nWe should either:\n\n1. Document that IDENTIFY_SYSTEM must always be run... |
[
{
"msg_contents": "Greetings\n\nConsider a table set up as follows:\n\n CREATE TABLE foo (id INT, val TEXT);\n ALTER TABLE foo DROP COLUMN val;\n\nWhen specifying the name of a dropped or non-existent column, the function\n\"has_column_privilege()\" works as expected:\n\n postgres=# SELECT has_column_p... |
[
{
"msg_contents": "Hi,\n\nWhile reviewing the patch in [1], I found that SearchCatCache function\nuses SearchCatCacheInternal as it's function name in the comment. It\nlooks like a typo, attaching a very small patch to correct it.\n\nThoughts?\n\n[1] - https://www.postgresql.org/message-id/CAD21AoAwxbd-zXXUAeJ2... |
[
{
"msg_contents": "Hi,\n\nThe function comment of RestoreBlockImage() seems not correct since it\nreturns a boolean, not the buffer number.\n\n /*\n * Restore a full-page image from a backup block attached to an XLOG record.\n *\n * Returns the buffer number containing the page.\n */\n bool\n RestoreBlockIm... |
[
{
"msg_contents": "Hi,\n\nWhile investigating a customer issue it's turned out that if a\nparallel worker executes inet_server_addr() and inet_server_port() the\nresults are always null because MyProcPort is not set in parallel\nworkers. We can reproduce it in all supported versions higher than\n9.6. Here is an... |
[
{
"msg_contents": "Although pg_publication_rel entries don't have an owner per se,\nit's still important for the ArchiveEntry created for one to be\nmarked with an appropriate owner, because that is what determines\nwhich role will be used to run the ALTER PUBLICATION command when\ndoing a restore with --use-se... |
[
{
"msg_contents": "I noticed that pg_dump --clean does not work with partitioned\nindexes. Given for instance\n\ncreate schema s1;\ncreate table s1.at1 (f1 int, f2 int, primary key(f1,f2)) partition by list(f1);\ncreate table s1.at11 partition of s1.at1 for values in(11);\ncreate table s1.at12 partition of s1.... |
[
{
"msg_contents": "Hi,\n\nThe parameter replication_timeout was retired in commit 6f60fdd701 back in\n2012, but some comments and error messages seem to refer to that old\nsetting instead of wal_sender_timeout or wal_receiver_timeout. The attached\npatch replaces the old language with more specific references.\... |
[
{
"msg_contents": "Hi\n\nIn function fetch_table_list, it get the table names from publicer and return a list of tablenames.\nWhen append the name to the list, it use the following code:\n\n**\n\tnspname = TextDatumGetCString(slot_getattr(slot, 1, &isnull));\n\tAssert(!isnull);\n\trelname = TextDatumGetCString(... |
[
{
"msg_contents": "Hello.\n\nCommit c6b92041d3 changed the definition of RelationNeedsWAL().\n\n-#define RelationNeedsWAL(relation) \\\n-\t((relation)->rd_rel->relpersistence == RELPERSISTENCE_PERMANENT)\n+#define RelationNeedsWAL(relation)\t\t\t\t\t\t\t\t\t\t\\\n+\t((relation)->rd_rel->relpersistence == RELPER... |
[
{
"msg_contents": "Hello,\n\nWe found problem that a huge amount of memory was consumed when\nwe created a foreign key on a partitioned table including a lots partitions\nand accessed them, as discussed in [1]. Kuroda-san's idea proposed in\nthis thread is reducing cached SPI plans by combining several plans i... |
[
{
"msg_contents": "Hi,\n\nin \"src/tutorial/funcs.source\" there is this block:\n\n\n-----------------------------\n-- Creating SQL Functions with multiple SQL statements\n-- you can also create functions that do more than just a SELECT.\n--\n-- 14MAR99 Clark Evans: Does not quite work, commented out for n... |
[
{
"msg_contents": "Hi team,\n\nI have a table with a column of \"timestamp without timezone\" and I want to\nalter it to a column of \"timestamp with timezone\" without rewriting the\nrows.\n\nLooking at the docs, and doing some quick tests, the data-on-disk is\nidentical for both columns. But when doing an alt... |
[
{
"msg_contents": "Hi,\n\nWhile reading the code about logical replication, I found that\nWalSndPrepareWrite function says it use XLogSendPhysical to fill out the\nsendtime, however, it actually done by WalSndWriteData. It looks like a\ntypo, attaching a very small patch to correct it.\n\n-- \nRegrads,\nJapin ... |
[
{
"msg_contents": "Hi all,\n\nI have just noticed that aef8948 has removed the last reference to\nPG_SHA256_DIGEST_STRING_LENGTH in the code (this was getting used to\nknow the length of a SHA256 digest encoded to hex for checksum\nmanifests, but pg_hex_enc_len() calculates the same in a more\nconsistent way). ... |
[
{
"msg_contents": "Hello,\n\n\nThe attached patch fixes a trivial mistake in the calculation of shmem TOC. The original code allocates unduly large memory because it adds the result of add_size() to its left argument.\n\n\nRegards\nTakayuki Tsunakawa",
"msg_date": "Thu, 14 Jan 2021 08:47:12 +0000",
"ms... |
[
{
"msg_contents": "Hello. I am interested in the question of profiling. As far as I\nunderstand, this issue has not been resolved in postgres. And I see\nthe urgency of this problem. Are there any postgres development plans\nfor this functionality?\n\nregards, Eremin Pavel.\n\n\n",
"msg_date": "Thu, 14 Jan ... |
[
{
"msg_contents": "Currently the cost_sort doesn't consider the number of columns to sort,\nwhich\nmeans the cost of SELECT * FROM t ORDER BY a; equals with the SELECT *\nFROM t ORDER BY a, b; which is obviously wrong. The impact of this is when\nwe\nchoose the plan for SELECT DISTINCT * FROM t ORDER BY c bet... |
[
{
"msg_contents": "As you may be aware the NOT VALID qualifier currently only applies to\nCHECK and FK constraints, but not yet to unique indexes. I have had\ncustomer requests to change that.\n\nIt's a reasonably common requirement to be able to change an index\nto/from a unique index, i.e. Unique -> NonUnique... |
[
{
"msg_contents": "Greetings\n\nConsider the following:\n\n postgres=# SELECT current_setting('server_version');\n current_setting\n -----------------\n 12.5\n (1 row)\n\n postgres=# SELECT * FROM pg_db_role_setting WHERE setrole = 10;\n setdatabase | setrole | setconfig\n ---... |
[
{
"msg_contents": "Hi\r\n\r\nI found a possible typo in reorderbuffer.c\r\n\r\n * has got a combocid. Combocid's are only valid for the duration of a\r\n\r\nCombocid's ==>> Combocids\r\n\r\nAttatching a small patch to correct it.\r\n\r\nBest regards,\r\nhouzj",
"msg_date": "Fri, 15 Jan 2021 04:58:34 +0000... |
[
{
"msg_contents": "Hello hackers,\n\nWhile following along with the nearby investigation into weird\ncross-version Apple toolchain issues that confuse configure, I noticed\nthat the newer buildfarm Macs say:\n\nchecking for fdatasync... (cached) yes\n\nThat's a bit strange because there's no man page and no dec... |
[
{
"msg_contents": "Hello hackers,\nWhen I do PITR in a strange step, I get this FATAL:\n2021-01-15 15:02:52.364 CST [14958] FATAL: hot standby is not possible because wal_level was not set to \"replica\" or higher on the primary server2021-01-15 15:02:52.364 CST [14958] HINT: Either set wal_level to \"replica... |
[
{
"msg_contents": "While reviewing parallel insert [1] (Insert into .... Select) and\nparallel copy patches [2], it came to my notice that both the patches\ntraverse the entire partition hierarchy to determine parallel-safety\nof partitioned relations. This is required because before considering\nthe Insert or ... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.