Using Compressed row_format in Percona xtradb cluster

MySQL & MariaDB Percona XtraDB Cluster 5.x
1

Hi
Because the tables are relatively large, we intend to convert existing tables to use compressed row_fromat in PXC.
This will reduce the amount of disk space used.

Please let us know if there are any known issues that need to be addressed when converting the row_format.

2

Hello @anjaneyarajus,
Converting tables to compressed will require complete table rebuilds. If you do this in TOI mode, it will block everything on the cluster for the duration. If you use RSU, you must alter each table on every node individually. You can also try NBO, which is like TOI but won’t block other operations, however not all ALTER types are supported with NBO. Of course, you can use TOI + pt-online-schema-change for best results.

3

HI Mathew,

are there any challenges in using compressed row_format?

4

No challenges. Compression is transparent to the app and queries.

5

Hi Mathew,

thank you for the response. we have observed the difference in disk space utilisation between a writer node and non-writer nodes.

the difference between writer node and non-writer is around 6%.
data space utilisation is more on non-writer nodes when compared to writer node. the data was generated using sysbench and used customised lua from one of the services.