My appologies. By 500K child nodes, I meant that this was a root, and
that the sum of the descendants were 500K. xxx/yyy has 5 children.

Most nodes have a small number (<20) of children. Part of the testing
was to see how I needed to change the data structure so that I
minimized the number of child nodes :) One of xxx/yyy's children was a
failed experiment with 1000 children.

Thanks for the answer I am very reassured: I can delete my failed
experiment nodes and I should be back working again. I'll tweak the
default number any way.

So is there a quick way of deleting a node which has hundred's of
thousands of descendants? Clicking delete in explorer causes a crash.
I can write a program to delete, but I suspect people have been here
before?


On Wed, Oct 19, 2011 at 9:19 PM, Justin Edelson
<[email protected]> wrote:
> Correction - the default maximum is 200, not 1000. Not sure what I was
> looking at there.
>
> Justin
>
> On Wed, Oct 19, 2011 at 4:18 PM, Justin Edelson
> <[email protected]> wrote:
>> Phil-
>> If the JSON object requested contains more than the maximum number of
>> nodes, an array is returned indicating which requests can be made to
>> return fewer than the max number of nodes.
>>
>> The typical use case for this is where /xxx contains /xxx/yy0 through
>> /yy9 and each of /xxx/yy0 through /xxx/yy9 contains 100+ child nodes.
>> If you requested /xxx.infinity.json, the result would contain 1012
>> nodes (if I'm doing the math right here). If that's more than the
>> allowed maximum, you'd get back ['/xxx.0.json', /xxx.1.json'] meaning
>> that either of those are requests which will be under the limit.
>>
>> This is a "feature" - http://issues.apache.org/jira/browse/SLING-1308
>>
>> In your case, if you really mean to say that you have 500k child nodes
>> under a single parent, you really should address this as that's not
>> going to be good for performance across the board.
>>
>> The maximum node count can be increased through the configuration of
>> the Default Get Servlet. The default is 1000.
>>
>> HTH,
>> Justin
>>
>> On Wed, Oct 19, 2011 at 3:24 PM, Phil Rice <[email protected]> wrote:
>>> I have a version of sling running that now has between 250K and 500K
>>> nodes underneath a node called "/xxx/yyy".
>>>
>>> When I do a get to  http://<host>/xxx/yyy/anything.1.json  I get a
>>> result that looks like:
>>> {"jcr:primaryType":"nt:unstructured","artifact":{"jcr:primaryType":"nt:unstructured"}}
>>>
>>> In this case artifact is the name of a node under /xxx/yyy/anything.
>>> As stated there are a lot of nodes under /xxx/yyy and my tests
>>> indicate that all of them return a Json map, and as far as I can tell
>>> the data matches the properties of the nodes.
>>>
>>> When I do a get to http://<host>/xxx/yyy.1.json I get a strange result
>>> back:  ["xxx/yyy.0.json"]
>>>
>>> When I do a get to http://<host>/xxx.1.json, I get the normal map.
>>>
>>> So of my half a million nodes, only this one is unusual. The nodes
>>> above it behave correctly, the nodes below it behave correctly. Only
>>> this one is strange.
>>>
>>> When I inspect the node /xxx/yyy with the explorer, it has only the
>>> default jcr:primaryType property.
>>>
>>> As further information the response to  http://<host>/xxx/yyy.0.json
>>> return a map: {"jcr:primaryType":"nt:unstructured"}
>>>
>>>
>>>
>>> What is even more disturbing is that while I was testing it, initially
>>> xxx/yyy.1.json behaved as normal, but about 2 days into the testing it
>>> changed. I have been able to reproduce this, although with a cycle
>>> time of many hours, it is hard to be certain what actions caused the
>>> behavior to change.
>>>
>>> I would appreciate any advice on:
>>> * What the result:  ["xxx/yyy.0.json"] actually means
>>> * Is this a bug, or is it a "feature" that I need to know about
>>>
>>> As this is on a server running live on the internet, I can pass ip
>>> address/port details and username/password details in private emails
>>> but I am loath to do that over an open email.
>>>
>>
>

Reply via email to