On Mon, 2019-02-25 at 18:53 -0500, marc.davenp...@oracle.com wrote: > I've been trying to rationalize the hit or miss nature of this > problem. > In my mind, I'm looking for a race condition. I'm going to bet that > this is a configuration issue on our end, but I'm not sure how this > would happen or where to look. Perhaps we are have some bundles out > of > order. While I had mentioned in a previous email that we had our > own > supplier of SlingRepository, I temporarily removed it trying to > debug > this issue. I can see that two different instances of > OakSlingRepositoryManager (which implement SlingRepository) are > created > on the first server startup. I saw this by setting a break point in > start(BundleContext, Config) of AbstractSlingRepositoryManager.
This reminds me of SLING-7811 [1]. While we did not get to a definitive solution, it might be worth upgrading the o.a.s.jcr.base bundle to 3.0.6 ( Sling 10 has 3.0.4 ) and see if the situation is improved. Thanks, Robert [1]: https://issues.apache.org/jira/browse/SLING-7811