Details
Description
Currently the replication list we returned to ns_server for ui is coming from our cache, but the other queries (via rest, cli) are directly against the metakv store. There are a chance of discrepancy among them.
We can't afford to go to metakv store for every query as it is expensive, so all the query should be against the cache. And any updates (update\add\delete) on replication spec would invalidate and repopulate the cache.
Attachments
Issue Links
- is duplicated by
-
MB-15175 goxdcr unable to remove remote cluster (windows)
- Closed
For Gerrit Dashboard: MB-14564 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
49727,17 | MB-14564 Query about replication specification should consistently be done against the same source -implement COW cache -use the cache for replication spec and remote cluster cache | master | goxdcr | Status: MERGED | +2 | +1 |