You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@geode.apache.org by Jens Deppe <je...@apache.org> on 2016/05/23 16:49:08 UTC

Index management

​Just recently I had reason to look​ at the gfsh *destroy index* call [1].
I noticed that an index can be destroyed using *any one* of the following
attributes: index name, region name, member name and group name.

Deleting an index by name or removing all indexes from a region makes
sense. However what does it mean to remove an index from a single member or
from a group? Wouldn't that result in the possibility of having an
'unbalanced' set of indexes where some members hosting a region have an
index defined, but other members don't? Seems like a recipe for serious
misconfiguration. I really don't understand the semantics of these two
scenarios. (I also couldn't find anything in the docs which explained this
more). Even the QueryService interface has no methods to delete indexes by
member or group.

Unless there is an obvious scenario I don't understand, I'd like to propose
we remove the ability (in gfsh at least) to destroy an index by *member* or
*group*.

--Jens

[1]
https://github.com/apache/incubator-geode/blob/8a3c351abd971ba740e77364a431aee7d5cc5b10/geode-core/src/main/java/com/gemstone/gemfire/management/internal/cli/commands/IndexCommands.java#L327

Re: Index management

Posted by Dan Smith <ds...@pivotal.io>.
+1 - You should only be able to remove an index from the whole region.

-Dan

On Mon, May 23, 2016 at 9:49 AM, Jens Deppe <je...@apache.org> wrote:

> ​Just recently I had reason to look​ at the gfsh *destroy index* call [1].
> I noticed that an index can be destroyed using *any one* of the following
> attributes: index name, region name, member name and group name.
>
> Deleting an index by name or removing all indexes from a region makes
> sense. However what does it mean to remove an index from a single member or
> from a group? Wouldn't that result in the possibility of having an
> 'unbalanced' set of indexes where some members hosting a region have an
> index defined, but other members don't? Seems like a recipe for serious
> misconfiguration. I really don't understand the semantics of these two
> scenarios. (I also couldn't find anything in the docs which explained this
> more). Even the QueryService interface has no methods to delete indexes by
> member or group.
>
> Unless there is an obvious scenario I don't understand, I'd like to propose
> we remove the ability (in gfsh at least) to destroy an index by *member* or
> *group*.
>
> --Jens
>
> [1]
>
> https://github.com/apache/incubator-geode/blob/8a3c351abd971ba740e77364a431aee7d5cc5b10/geode-core/src/main/java/com/gemstone/gemfire/management/internal/cli/commands/IndexCommands.java#L327
>

Re: Index management

Posted by Anilkumar Gingade <ag...@pivotal.io>.
Agree....

The indexes are operated using api/xml at a peer/server node level; there
is no client-server index operation command; one reason index destroy is
supported on member, to simulate what api is supported...As you mentioned
it may not make sense to have it from Gfsh.

-Anil.


On Mon, May 23, 2016 at 9:49 AM, Jens Deppe <je...@apache.org> wrote:

> ​Just recently I had reason to look​ at the gfsh *destroy index* call [1].
> I noticed that an index can be destroyed using *any one* of the following
> attributes: index name, region name, member name and group name.
>
> Deleting an index by name or removing all indexes from a region makes
> sense. However what does it mean to remove an index from a single member or
> from a group? Wouldn't that result in the possibility of having an
> 'unbalanced' set of indexes where some members hosting a region have an
> index defined, but other members don't? Seems like a recipe for serious
> misconfiguration. I really don't understand the semantics of these two
> scenarios. (I also couldn't find anything in the docs which explained this
> more). Even the QueryService interface has no methods to delete indexes by
> member or group.
>
> Unless there is an obvious scenario I don't understand, I'd like to propose
> we remove the ability (in gfsh at least) to destroy an index by *member* or
> *group*.
>
> --Jens
>
> [1]
>
> https://github.com/apache/incubator-geode/blob/8a3c351abd971ba740e77364a431aee7d5cc5b10/geode-core/src/main/java/com/gemstone/gemfire/management/internal/cli/commands/IndexCommands.java#L327
>