Hello!
I will give an alternative explanation.
Even if Ignite will not go to the underlying database when doing its own
SQL, it has to use underlying database to do readThrough, so underlying
database should have an index on the cache key.
If you only ever do writeThrough, I think you can go without any keys or
indexes in some extreme cases.
Regards,
--
Ilya Kasnacheev
ср, 4 дек. 2019 г. в 00:52, Stefan Miklosovic <
stefan.mikloso...@instaclustr.com>:
> Hi,
>
> I am using Cassandra Ignite integration. In this integration, one is
> able to provide specific "mapping" in persistence xml so a key is e.g.
> a primitive and value is POJO.
>
> Now, I was quite surprised (as a lot of other people too as I read
> more about this), that it is not possible to do SQL queries on data
> which are not in memory / cached as a query is not going down to DB
> level when cache is empty - one has to load it all into memory or one
> has to know a key so it will be cached (and sql queries can work on
> it).
>
> Do not get me wrong but once this is the reality, what is the purpose
> of secondary indexes in Cassandra table? For example, I have this
> POJO:
>
> @Builder
> @Getter
> @Setter
> @ToString
> @EqualsAndHashCode
> @AllArgsConstructor
> @NoArgsConstructor
> public class CustomCounterIgnite implements Serializable {
>
> @NotNull
> @QuerySqlField(index = true)
> private String username;
>
> @NotNull
> @QuerySqlField(index = true)
> private Integer counterone;
>
> @NotNull
> @QuerySqlField(index = true)
> private Integer countertwo;
> }
>
> It is said that "in order to do some selects / filtering on fields,
> you have to set "index = true" for these fields".
>
> But, given how this SQL mechanism works, once it is in memory, cached,
> why there are indexes created in database? What is the purpose if I
> can do SQL queries only on data in memory? It does not matter if some
> indexes are created in Cassandra or not because it will never go to
> database ....
>
> On initialisation, this is the output:
>
> create table if not exists "ignite"."custom_counter"
> (
> "username" text,
> "countertwo" int,
> "counterone" int,
> primary key (("username"))
> );
>
> Creating indexes for Cassandra table 'ignite.custom_counter'
> create index if not exists on "ignite"."custom_counter" ("countertwo");
> create index if not exists on "ignite"."custom_counter" ("counterone");
> Indexes for Cassandra table 'ignite.custom_counter' were successfully
> created
>
> Thanks!
>
