Open paterczm opened 7 years ago
It's easy to miss the fact that one of the indexes wasn't created, since metadata is updated with both indexes. The only indication that something may be amiss is lack of successful response from metadata svc (since it times out).
We can move the index creation into its own thread, and return immediately.
On Thu, Apr 6, 2017 at 9:26 AM, Marek notifications@github.com wrote:
It's easy to miss the fact that one of the indexes wasn't created, since metadata is updated with both indexes. The only indication that something may be amiss is lack of successful response from metadata svc (since it times out).
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/lightblue-platform/lightblue-rest/issues/285#issuecomment-292210166, or mute the thread https://github.com/notifications/unsubscribe-auth/ADgDDe-v7ZW4sGYxcdmDVzzvhmi64tegks5rtQQ8gaJpZM4M1wVK .
Yes, for indexes with background: true that makes sense.
This issue may belong to lightblue-mongo. I decided to put it here since it has to do with http timeouts.
Recently, I was pushing 2 new indexes on a large collection, which resulted in the metadata update request to timeout. Each time I did the push only the first index was created. I had to repeat the metadata update request to create the other one.