Thread: Reindex doesn’t not working replica nodes
Dear experts,
We have initiated reindex on source db but it not replicated to replica db and we have having select query fetching with zero rows but data exiting on the table.
Postgres 11.10 streaming replication
Postgres 11.10 streaming replication
Regards,
Mohammed Afsar
Hi Mohammed,
PostgreSQL replication is based on replicating WAL (Write-Ahead Log) records, which capture changes to data files such as inserts, updates, deletes, and the creation of new tables. However,
WAL does not include the physical structure of indexes.
If you're encountering issues like zero rows being fetched from the replica, it could be due to index inconsistencies or replication lag. To ensure your read replica has updated indexes, you can promote the replica to be the primary node, perform the REINDEX operation, and then revert the roles, making the original primary node the primary again.
Regards
Muhammad Affan
PostgreSQL replication is based on replicating WAL (Write-Ahead Log) records, which capture changes to data files such as inserts, updates, deletes, and the creation of new tables. However,
WAL does not include the physical structure of indexes.
If you're encountering issues like zero rows being fetched from the replica, it could be due to index inconsistencies or replication lag. To ensure your read replica has updated indexes, you can promote the replica to be the primary node, perform the REINDEX operation, and then revert the roles, making the original primary node the primary again.
Regards
Muhammad Affan
On Wed, Aug 14, 2024 at 5:58 PM Mohammed Afsar <vmdapsar@gmail.com> wrote:
Dear experts,We have initiated reindex on source db but it not replicated to replica db and we have having select query fetching with zero rows but data exiting on the table.
Postgres 11.10 streaming replicationRegards,Mohammed Afsar
Thanks for the information but we are encountered OS compatibility issues source to replica servers.
Source Alma Linux 9.0
Replica centos 7.0
Patroni 3.1.0
Regards,
Mohammed Afsar
On Fri, 16 Aug 2024 at 4:50 PM, khan Affan <bawag773@gmail.com> wrote:
Hi Mohammed,
PostgreSQL replication is based on replicating WAL (Write-Ahead Log) records, which capture changes to data files such as inserts, updates, deletes, and the creation of new tables. However,
WAL does not include the physical structure of indexes.
If you're encountering issues like zero rows being fetched from the replica, it could be due to index inconsistencies or replication lag. To ensure your read replica has updated indexes, you can promote the replica to be the primary node, perform the REINDEX operation, and then revert the roles, making the original primary node the primary again.
Regards
Muhammad AffanOn Wed, Aug 14, 2024 at 5:58 PM Mohammed Afsar <vmdapsar@gmail.com> wrote:Dear experts,We have initiated reindex on source db but it not replicated to replica db and we have having select query fetching with zero rows but data exiting on the table.
Postgres 11.10 streaming replicationRegards,Mohammed Afsar
Hi Mohammed
If we are talking physical replication then the wall includes changes to index’s data. If we are talking logical replication the wall does not include index changes.
Med venlig hilsen
Peter Gram
Sæbyholmsvej 18
Peter Gram
Sæbyholmsvej 18
2500 Valby
Mobile: (+45) 5374 7107
Email: peter.m.gram@gmail.com
On Fri, 16 Aug 2024 at 13.20, khan Affan <bawag773@gmail.com> wrote:
Hi Mohammed,
PostgreSQL replication is based on replicating WAL (Write-Ahead Log) records, which capture changes to data files such as inserts, updates, deletes, and the creation of new tables. However,
WAL does not include the physical structure of indexes.
If you're encountering issues like zero rows being fetched from the replica, it could be due to index inconsistencies or replication lag. To ensure your read replica has updated indexes, you can promote the replica to be the primary node, perform the REINDEX operation, and then revert the roles, making the original primary node the primary again.
Regards
Muhammad AffanOn Wed, Aug 14, 2024 at 5:58 PM Mohammed Afsar <vmdapsar@gmail.com> wrote:Dear experts,We have initiated reindex on source db but it not replicated to replica db and we have having select query fetching with zero rows but data exiting on the table.
Postgres 11.10 streaming replicationRegards,Mohammed Afsar
Thanks for sharing the additional information. Initially, the query didn't provide the full context regarding OS compatibility, which is crucial in understanding the replication issues (physical and logical).
For future inquiries, including details like OS versions, replication setups, and any specific configurations can help us address your concerns more accurately and efficiently.
On Fri, Aug 16, 2024 at 5:19 PM Mohammed Afsar <vmdapsar@gmail.com> wrote:
Thanks for the information but we are encountered OS compatibility issues source to replica servers.Source Alma Linux 9.0Replica centos 7.0Patroni 3.1.0Regards,Mohammed AfsarOn Fri, 16 Aug 2024 at 4:50 PM, khan Affan <bawag773@gmail.com> wrote:Hi Mohammed,
PostgreSQL replication is based on replicating WAL (Write-Ahead Log) records, which capture changes to data files such as inserts, updates, deletes, and the creation of new tables. However,
WAL does not include the physical structure of indexes.
If you're encountering issues like zero rows being fetched from the replica, it could be due to index inconsistencies or replication lag. To ensure your read replica has updated indexes, you can promote the replica to be the primary node, perform the REINDEX operation, and then revert the roles, making the original primary node the primary again.
Regards
Muhammad AffanOn Wed, Aug 14, 2024 at 5:58 PM Mohammed Afsar <vmdapsar@gmail.com> wrote:Dear experts,We have initiated reindex on source db but it not replicated to replica db and we have having select query fetching with zero rows but data exiting on the table.
Postgres 11.10 streaming replicationRegards,Mohammed Afsar
Hi Peter,
We are using streaming replications but still reindex is not working source db has 11.15replica has 11.11 PostgreSQL version.
Regards,
Mohammed Afsar
On Fri, 16 Aug 2024 at 6:33 PM, Peter Gram <peter.m.gram@gmail.com> wrote:
Hi MohammedIf we are talking physical replication then the wall includes changes to index’s data. If we are talking logical replication the wall does not include index changes.Med venlig hilsen
Peter Gram
Sæbyholmsvej 182500 ValbyMobile: (+45) 5374 7107Email: peter.m.gram@gmail.comOn Fri, 16 Aug 2024 at 13.20, khan Affan <bawag773@gmail.com> wrote:Hi Mohammed,
PostgreSQL replication is based on replicating WAL (Write-Ahead Log) records, which capture changes to data files such as inserts, updates, deletes, and the creation of new tables. However,
WAL does not include the physical structure of indexes.
If you're encountering issues like zero rows being fetched from the replica, it could be due to index inconsistencies or replication lag. To ensure your read replica has updated indexes, you can promote the replica to be the primary node, perform the REINDEX operation, and then revert the roles, making the original primary node the primary again.
Regards
Muhammad AffanOn Wed, Aug 14, 2024 at 5:58 PM Mohammed Afsar <vmdapsar@gmail.com> wrote:Dear experts,We have initiated reindex on source db but it not replicated to replica db and we have having select query fetching with zero rows but data exiting on the table.
Postgres 11.10 streaming replicationRegards,Mohammed Afsar
> On Aug 16, 2024, at 7:08 AM, Mohammed Afsar <vmdapsar@gmail.com> wrote: > > We are using streaming replications but still reindex is not working source db has 11.15replica has 11.11 PostgreSQL version. This is not your problem. As you understand, I think, with streaming replication all files, including indexes, are exactly duplicated on the replica.If the OS on the two sides have different glibc versions which have different collations, then the index structureof the primary will be a corrupted index on the replica. You have to fix your OS version incompatibility, or moveto a version of postgres which can use ICU collations and make sure that library is compatible on both sides.
Got it thanks for the deep dive explanation.Scott 👍🏻
And parallel we got below error also included OS incompatibility?
postgres[265938]: segfault at 18 ip 00000000004cf11b
Regards,
Mohammed Afsar
On Fri, 16 Aug 2024 at 6:49 PM, Scott Ribe <scott_ribe@elevated-dev.com> wrote:
> On Aug 16, 2024, at 7:08 AM, Mohammed Afsar <vmdapsar@gmail.com> wrote:
>
> We are using streaming replications but still reindex is not working source db has 11.15replica has 11.11 PostgreSQL version.
This is not your problem.
As you understand, I think, with streaming replication all files, including indexes, are exactly duplicated on the replica. If the OS on the two sides have different glibc versions which have different collations, then the index structure of the primary will be a corrupted index on the replica. You have to fix your OS version incompatibility, or move to a version of postgres which can use ICU collations and make sure that library is compatible on both sides.