Slack channels indexing error Unfenced index attempt found in DB

[plopezamaya]

Hello ✋ ,

We’ve noticed that the slack channels can have some troubles indexing when there are multiple slack channels being indexed at the same time. Probably due to the following :

• Each slack channel being indexed have to list all conversation before finding the the new conversations to index. This causes a rate limit on the api and at some point the indexing just fails because of the session time out in the database.

• Maybe I'm wrong but it seems the date it is starting to get the conversations is not correct. As for a single document it takes a lot of time to find it (3000 seconds for a single new conversation in one slack channel)

• Usually for confluence there is only one indexing at a time. With slack there are multiple indexing at a time, plus de parallel indexing of the other connectors (e.g confluence). Vespa timeout as it doesn’t follow the load of the multiple channels indexing of slack, plus the other connectors. It would be better for vespa workload to only index one connector "space" at a time.

It starts indexing again once we restart all the background and/or api-server containers but then it fails because of the explained issues.

We are currently on the v0.16.6