Important addition!!!!!

The App just stopped working again and is giving me the NeedIndexError again! This
coincided with my re-addition of the superset (ancestor only) index going from status
"building" to "serving".

So, I think it might be clear what is going on:

With an index setup like this:

  indexes:

  - kind: Segment
    ancestor: yes

  - kind: Segment
    ancestor: yes
    properties:
    - name: rev

The second index seems to be hidden by the first one, since it is a subset of the
first. Deleting the first index exposes the second one.

After deleting the first index out of index.yaml and running "appcfg.py update", a
call to "appcfg.py vacuum_indexes" outputs this:

  Fetching index definitions diff.
  This index is no longer defined in your index.yaml file.

  ancestor: true
  kind: Segment
  properties: []

  Are you sure you want to delete this index? (N/y/a): y
  Deleting selected index definitions.
  Error 500: --- begin server output ---

  Server Error (500)
  A server error has occured.
  --- end server output ---

After this, the app works again, despite the error.

This problem only seems to be happening if an ancestor is involved.

I hope this helps.

Thanks,

  Markus

I've repro'd Markus's problem in a simpler context, which may help with tracking it
down.  (BTW Markus, bigsheetofpaper.com is *great* :) )

After completing the App Engine tutorial, I ran appcfg update.  Attempting to view
the root URL [which requests Greeting.all().order('-date').fetch(10)] gave a
NeedIndexError for the following index spec:
 
  - kind: Greeting
    properties:
    - name: date
      direction: desc

This was the content of my index.yaml file, minus comments:

  - kind: Greeting
    properties:
    - name: author
    - name: date
      direction: desc

  - kind: Greeting
    properties:
    - name: date
      direction: desc

Both indexes showed as "serving" on my App dashboard's Index page.

Taking a hint from Markus, I deleted the first entry in my index.yaml file, which was
an unused remnant from earlier code.  After another appcfg update, I no longer
received the error, and only one index is in my dashboard's list.

Interestingly, I then put back the deleted entry and re-updated, and the error does
not reappear.  Both indexes are listed again on the dashboard.

Hope this helps,
Michael [gundlach at you-can-guess-where]

Just wait...
when the Data Store Indexes' status is "Serving", I think everything will be ok.
Hope this is the reason.
Actually, I don't know how it works...

(No comment was entered for this change.)

I have the same problem. In my app I have two indices, one is a superset of another (has an 
additional property):

# index A
- kind: UserRelease
  ancestor: yes
  properties:
  - name: artist_id
  - name: order
    direction: desc

# index B
- kind: UserRelease
  ancestor: yes
  properties:
  - name: order
    direction: desc

The query involving index A raises a NeedIndexError exception even though the index is available and 
its status is "Serving". 

If I vacuum index B, index A starts to work. When I put back index B, index A doesn't work again. 

The error occurs only in the production environment, everything works fine locally. It looks like 
index B somehow "hides" index A. This is clearly a bug, please fix it.

Forgot to mention. I didn't get this error until I started using ancestor indices. 
That is, if in the indices above I replace "ancestor: yes" with an additional 
property (e.g. "- name: user_id") and update the entities and queries accordingly, 
then both indices work fine.

I am also having this issue, also using ancestor. For me, it doesn't seem to matter
which order the indexes are in index.yaml -- whichever index definition comes first
"blocks" the other index definition, so only one can be used.

This is especially frustrating because I just refactored a bunch of code to make use
of the ancestor relationship. Meh.

(No comment was entered for this change.)

This one really needs to be addressed...  

Fixed in 1.1.6